The Translocation Domain in Trimeric Autotransporter Adhesins Is Necessary and Sufficient for Trimerization and Autotransportation

Abstract

Trimeric autotransporter adhesins (TAAs) comprise one of the secretion pathways of the type V secretion system. The mechanism of their translocation across the outer membrane remains unclear, but it most probably occurs by the formation of a hairpin inside the β-barrel translocation unit, leading to transportation of the passenger domain from the C terminus to the N terminus through the lumen of the β-barrel. We further investigated the phenomenon of autotransportation and the rules that govern it. We showed by coexpressing different Escherichia coli immunoglobulin-binding (Eib) proteins that highly similar TAAs could form stochastically mixed structures (heterotrimers). We further investigated this phenomenon by coexpressing two more distantly related TAAs, EibA and YadA. These, however, did not form heterotrimers; indeed, coexpression was lethal to the cells, leading to elimination of one or another of the genes. However, substituting in either protein the barrel of the other one so that the barrels were identical led to formation of heterotrimers as for Eibs. Our work shows that trimerization of the β-barrel, but not the passenger domain, is necessary and sufficient for TAA secretion while the passenger domain is not.

INTRODUCTION

The type V secretion system is the most widespread mechanism of protein secretion in disease-causing Gram-negative bacteria (6). It is a Sec-dependent system consisting of three distinct pathways: the classical (monomeric) autotransporter pathway (type Va), two-partner secretion (type Vb), and the trimeric autotransporter pathway (type Vc).

Classical autotransporters are single-chain proteins comprised of a signal peptide, a passenger domain that is exposed to the extracellular space and is in many cases cleaved after transportation, and a C-terminal translocation domain. Passenger domains contain the specific activity of the protein (14) and usually have a central β-helical core (9, 11, 31) though there are exceptions (44). The C-terminal translocation domain forms a 12-stranded (30) β-barrel that is responsible for insertion in the outer membrane (OM) and translocation of the passenger (32). Trimeric autotransporter adhesins (TAAs), which comprise the type Vc secretion pathway (26), are built of three identical polypeptide chains carrying a signal peptide that directs them to the Sec-machinery. As for classical autotransporters, the structure of TAAs can be divided into distinct regions: the N-terminal passenger domain typically contains a trimeric lollipop-like head, a neck, and a coiled-coil stalk domain (15), followed by a C-terminal translocation domain that is, as for classical autotransporters, believed to be responsible for insertion and translocation of the passenger domain to the outside of the cell (Fig. 1E). The translocation domain is a 12-stranded β-barrel in which four strands come from one monomeric polypeptide (27).

Fig 1.

Models of passenger domain translocation across the OM. (A) Protein with folded β-barrel and inserted into the OM, before translocation and folding of the passenger domain. Proposed models for passenger translocation are also shown: threading model (a submodel of the single-chain model), in which the N terminus is translocated first and then the rest of the protein is translocated with an N-to-C polarity (B); hairpin model (second submodel of the single-chain models), where the C terminus of the passenger domain is first inserted into the pore of the β-barrel so that a hairpin structure is created and the rest of the protein is translocated C to N (C); and a Bam complex-assisted model in which the Bam complex is involved not only in β-barrel insertion into the OM but also in passenger domain translocation (D). Panel E presents the folded structure of EibD, with protein domains marked on the right-hand side (24).

In autotransporter systems, proteins are transported across the inner and outer membranes in two distinct steps. The first step for all types is translocation across the inner membrane in a Sec-dependent manner (6), directly followed by passage through the periplasm. During this step, an important role is attributed to the signal peptide, which acts as a transient membrane tether maintaining the passenger domain in an unfolded state and thus allowing passage of the protein through the periplasm (42).

The second step, insertion into the outer membrane and translocation to the outside of the cell, is different for both pathways. However, the assistance of the Bam complex (previously known as Omp85 or YaeT) is required in both (16, 20, 22, 25, 40). Strains with deletion of Omp85 were not able to transport IgA protease, a classical autotransporter. Its passenger domain, which is cleaved after transportation on the cell surface, accumulated in a full-length form in the periplasm (46). Recently, Hagan et al. (13) showed in an ex vivo experiment that the Bam complex assembles β-barrel proteins into the OM.

Much work has been done to elucidate the mechanism of autotransportation and folding in classical autotransporters (18, 19, 29, 30, 34, 39, 45). Of the proposed models to explain autotransportation, two are still considered viable, the single-chain model and the Bam complex-assisted model. In both, the C-terminal domain folds first into the OM with the assistance of the Bam complex.

The single-chain model posits that the passenger domain is transported through the pore of the barrel to the extracellular space without the assistance of other proteins, ending with an N-out topology, where the N terminus is distal to the OM and the C terminus is on the inside (2). The way in which the passenger domain is transported remains under debate. Does the N terminus of the passenger domain thread through the pore of the barrel, or do the C terminus of the passenger domain and the N terminus of the barrel form a hairpin structure, leading to transport of the passenger domain (30)? Considerable evidence supports the hairpin model: the presence of folding promoting regions in BrkA (29); mutagenesis of the linking region, which affects passenger translocation but not barrel insertion (21); and the fact that C-terminal disulfide bridges in the Prn passenger domain prevent surface exposure, whereas N-terminal ones do not (17).

In the second model, the Bam complex is posited to stabilize an open conformation of the β-barrel that could allow partially folded passenger domain passage through the pore (30). Recent results conducted with OmpT and purified Bam complex in proteoliposomes indicate that insertion and folding of outer membrane proteins occurs without any energy source; i.e., the process is driven by the ΔG of protein folding (13). Moreover, it has also been postulated by Ieva and Bernstein (16) that the Bam complex is involved in passenger assembly. Recent results published by Sauri et al. showed that the β-barrel is necessary for both insertion into the OM and translocation of the passenger domain and that it cooperates with BamA complex (39). The topology of the transport through the Bam complex-assisted model and the single-chain model can be the same; passage through the pore stabilized by the Bam complex may still occur either from the N terminus to the C terminus or from the C terminus to the N terminus.

The mechanism of autotransportation of TAAs has been less well-studied but is generally assumed to be similar to the classical autotransporters (4) (Fig. 1). The most probable mechanism of autotransportation is analogous to the single-chain model, but it is still unclear whether the N or the C terminus of the passenger domain is first transported through the pore of the β-barrel. The experimental data support the hairpin model, where three hairpins are formed from the three passenger domains (Fig. 1C). First, the entire passenger domain is not required for transportation (27, 35). Second, trimerization of the passenger domain seems to occur after trimerization and incorporation of the translocation domain into the outer membrane (5). Third, some TAA passenger domains are over 3,000 residues in length (43), making threading of the three N termini unlikely. On the other hand, the space available inside the β-barrel in its final conformation is limited, so even though three β-hairpins (with each β-strand approximately 4 Å in width) could fit into the barrel, the space would be very tight (27). Another possibility would therefore be that the C-terminal part of the hairpin—possibly as an already folded helix as in the NalP structure (30)—is already in the pore, while the N-terminal part is outside, surrounded by the Bam complex (Fig. 1D). This is not consistent with recent results from Leyton et al. (25) on monomeric ATs, suggesting that the role of the Bam complex is to keep the transporter domain in a transportation-competent state but that there are still very strict constraints on what can pass through the barrel. It is also hard to reconcile this structure with the absolute requirement for Gly in the center of the barrel (12).

Our study addresses the mechanism of translocation. We used two different types of TAAs, YadA expressed by enteropathogenic Yersinia strains and Eib proteins (EibA, EibC, and EibD) from Escherichia coli reference strains, to study the role of the translocation domain in transport and folding. We showed that two different Eibs can form heterotrimers but that YadA does not form heterotrimers with EibA. Moreover, we showed that, though YadA/EibA heterotrimer does not form, it will when the barrel of YadA is replaced by the EibA β-barrel or vice versa. Taken together, our work shows that a functional β-barrel is necessary and sufficient for trimerization and autotransportation.

MATERIALS AND METHODS

Cloning of YadA and Eibs for cross-trimerization experiments.

We first constructed a novel vector for coexpression of two different membrane proteins on the cell surface so that expression of the two proteins is induced by isopropyl-β-d-thiogalactopyranoside (IPTG), with each having a unique signal sequence and tag. We used different signal sequences to increase the genetic stability of the plasmid in expression strains by avoiding repeated sequences. We modified the expression vector pETDuet-1 (Novagen), which allows expression of two inserts from separate T7 promoters, to include signal sequences for periplasmic targeting. We amplified the signal sequence of EibD by PCR and inserted this between the NdeI and BglII sites of the second polylinker region of pETDuet-1 (primer sequences in are given in Table S1 in the supplemental material). The primers were designed to include an N-terminal StrepII tag, located directly after the cleavage site for the signal peptide. We then isolated the PelB signal sequence from pET22b (Novagen) by PCR (using T7 and T7 terminator primers) and cloned the resulting product between the XbaI and NcoI sites in the first polylinker region of pETDuet-1. This resulted in the first multiple cloning site that included a hexahistidine tag directly after the PelB cleavage site. Thus, this vector can be used to express two inserts with signal peptides; the first will have an N-terminal His tag, and the second will have an N-terminal StrepII tag. We called this vector pETDuet-S (for secretion), in which the products of both inserts are targeted to the periplasm by the Sec machinery.

For the cloning of YadA and of EibA, EibC, and EibD, we amplified the sequences of the passenger domains and translocation units of the corresponding genes, omitting the native signal sequences (see Table S1 in the supplemental material). The products were then cloned into the NcoI/HindIII sites (EibD), BamHI/HindIII sites (YadA), or BglII/AatII sites (EibA and EibC), either alone or in combination (Table 1).The correctness of all insertions was verified by sequencing.

Table 1.

Plasmids and constructs used in our studies with description of tags used to detect heterotrimers

Plasmid	Insert(s)a (cloning site[s])	Comment
pETDuet-S		Derivative of pETDuet-1; contains signal sequences for periplasmic targeting of both inserts
pETDuetS-EibA	eibA29-392 (BglII, AatII)	Contains an N-terminal Strep II tag
pETDuetS-EibAD	eibA29-392 (BglII, AatII), eibD27-511 (NcoI, HindIII)	EibA contains an N-terminal Strep II tag; EibD contains an N-terminal His tag
pETDuetS-EibD	eibD27-511 (NcoI, HindIII)	Contains an N-terminal His tag
pETDuetS-EibC	eibC26-504 (BglII, AatII)	Contains an N-terminal Strep II tag
pETDuetS-EibCD	eibC26-392 (BglII, AatII), eibD27-511 (NcoI, HindIII)	EibC contains an N-terminal Strep II tag; EibD contains an N-terminal His tag
pETDuetS-YadA	yadA26-455 (BamHI, HindIII)	Contains an N-terminal His tag
pETDuetS-EibA-YadA	eibA29-298-yadA368-455 (BglII, AatII)	Contains an N-terminal Strep II tag
pETDuetS-YadA/EibA-YadA	yadA26-455 (BamHI, HindIII), eibA29-298-yadA368-455 (BglII, AatII)	YadA contains an N-terminal His tag; the fusion of the EibA passenger and YadA translocator domains contains an N-terminal Strep II tag
pETDuetS-Yada-EibA	yadA26-367-eibA299-392 (BamHI, HindIII)	Contains N-terminal His tag
pETDuetS-EibA/YadA-EibA	yadA26-367-eibA299-392 (BamHI, HindIII), eibA29-392 (BglII, AatII)	Fusion of the YadA passenger and EibA translocator domains contains an N-terminal His tag; EibA contains an N-terminal Strep II tag
pETDuetS-YadA/EibA	eibA29-392 (BglII, AatII), yadA26-455 (BamHI, HindIII)	EibA contains an N-terminal Strep II tag; YadA contains an N-terminal His tag

^aThe subscript after the gene name refers to the amino acid residues of the corresponding protein encoded by the cloned sequence.

We produced the YadA-EibA fusions by a two-step protocol. We separately amplified the sequences corresponding to the YadA and EibA passenger domains and translocation units. The forward primers of the translocation units were designed to contain a 5′ overhang complementary to the 3′ end of the passenger domain to be fused (see Table S1). We then annealed the products of these reactions (YadA passenger domain to the EibA barrel, forming YadA-EibA, and vice versa, forming EibA-YadA) and filled in the fragments by PCR. The resulting inserts were then digested with BamHI and HindIII (YadA-EibA) or BglII and AatII (EibA-YadA). These were then cloned into pETDuet-S, either alone or in combination with wild-type YadA or EibA.

Expression and outer membrane protein purification.

We transformed the plasmids for surface expression of full-length Eibs and YadA into the strain BL21(DE3) Omp8, which is optimized for the expression of OM proteins (33). These bacteria were grown overnight at 37°C in 5 ml of Luria-Bertani (LB) medium supplemented with ampicillin at 100 μg/ml. The following morning the optical density (OD) was measured, and the bacteria were diluted to an OD of ∼0.3 in 20 ml of fresh LB supplemented with 100 μg/ml ampicillin and allowed to grow to mid-log phase (OD at 600 nm [OD₆₀₀] of ∼0.5). Protein production was then induced by the addition of 0.2 mM IPTG. After a further 2 h of incubation at 37°C, we diluted the cells to an OD₆₀₀ of 0.6. Cells from 10 ml of this suspension were harvested by centrifugation for 10 min at 4,000 × g. We then washed the cells by resuspending them in 1.5 ml of 10 mM HEPES (pH 7.4), centrifuged them again for 10 min at 4,000 × g, and resuspended the pellet in 2 ml of 10 mM HEPES (pH 7.4) for OM isolation. Cell lysates were produced by sonicating the resuspended cells twice for 30 s each time (MS73 probe [Bandelin Sonoplus]; 0.5-s duty cycle) and subjecting them to centrifugation for 2 min at 15,600 × g to remove cell debris. To isolate the OM, we followed the protocol of Carlone et al. (3). Briefly, inner membranes were separated from Sarkosyl-insoluble OMs by solubilization with 400 μl of 1% N-lauroyl sarcosine. Pelleted, washed OMs were resuspended in 50 μl of 10 mM HEPES (pH 7.4) and solubilized in SDS-containing sample buffer. They were then used immediately and run on SDS-PAGE gels or stored at −20°C.

Nondissociating SDS-PAGE.

Electrophoresis was performed with samples containing nonreducing SDS-PAGE sample buffer (41). Before loading, samples were heated at 50°C for 10 min and then loaded on the ready-made 4 to 12% gradient gel (Bio-Rad Mini-Protean Precast Gels), and electrophoresis was performed at 160 V. These conditions prevent the very stable TAA trimers from dissociating.

Proteinase K treatment of TAA-expressing cells.

Bacteria were cultured and induced as above. Cultures after induction were diluted to an OD₆₀₀ of 0.6 in a volume of 20 ml and collected by centrifugation for 10 min at 4,000 × g. The pellet was resuspended in 4 ml of phosphate-buffered saline (PBS) with the addition of 25 μg/ml of chloramphenicol to inhibit protein synthesis. After 5 min of incubation at room temperature (RT), the samples were split in half. To one half we added proteinase K (to a final concentration of 40 U/ml) and incubated the samples for 15 min at RT, while the other half were used as controls and treated in all respects the same in the subsequent steps. Proteolysis was stopped by addition of phenylmethylsulfonyl fluoride (PMSF) to a final concentration of 0.2 mM to both proteinase K-treated and control cells. The cells were pelleted by centrifugation for 10 min at 4,000 × g, resuspended in 1.5 ml of cold (4°C) 10 mM HEPES, pH 7.4 (containing protease inhibitor as above), recentrifuged for 10 min at 4°C at 4,000 × g, and resuspended in 2 ml of cold 10 mM HEPES, pH 7.2 (containing protease inhibitor as above). From that point samples were either treated as in the protocol for outer membrane purification (above) or stored at −80°C for subsequent use.

Western blotting.

After SDS-PAGE (described above), the protein samples were blotted onto nitrocellulose filters (Hybond C extra; GE Healthcare) with a Bio-Rad semidry apparatus, according to the manufacturer's instructions using standard transfer buffer. Blocking and incubations with antibodies differ for the various blots and are described below. The final step, detection of the tagged protein, was again the same for all steps. Proteins were detected by enhanced chemiluminescence (ECL Plus kit; GE Healthcare) according to the manufacturer's instructions.

Detection of StrepII-tagged proteins using StrepTactin conjugated to HRP.

The membrane was blocked for 1 h at RT in 1% fat-free milk powder in 20 ml of PBS with 0.1% (vol/vol) Tween. Then, we washed membrane three times for 5 min in 20 ml of PBS–0.1% Tween, and this step was followed by incubation in 10 ml of PBS–0.1% Tween with 5 μl of StrepTactin conjugated to horseradish peroxidase ([HRP] diluted before use according to the manufacturer's protocol). The membrane was then washed two times in 20 ml of PBS–0.1% Tween for 5 min each and two times in 20 ml of PBS for 5 min each. Finally, the Strep-tagged protein was detected (see above).

Detection of His-tagged proteins using a Ni-NTA conjugate.

The steps after the Western blotting procedure were done according to the manufacturer's protocol (HisDetector Nickel-HRP, catalog number 24-01-01; KPL). Briefly, membrane was blocked in 20 ml of 1% bovine serum albumin (BSA) at RT for 1 h, and then a Ni-nitrilotriacetic acid (NTA)-HRP conjugate was added at a 1:20,000 dilution. After incubation for 1 h at RT, the membrane was washed three times for 5 min each time in 20 ml of 1× Tris-buffered saline–0.1% Tween (TBST), and proteins were detected (see above).

IgG Fc binding assay.

After the membrane was blotted, it was blocked for 1 h in 20 ml of PBS–1% fat free milk powder at RT. Human IgG Fc-HRP (Jackson ImmunoResearch) diluted 1:8,000 was added to the blocking solution, and the sample was again incubated for 1 h in RT, washed in 20 ml of PBS three times for 5 min each time, and then detected (see above).

Collagen far-Western blotting.

After the blotting step, the membrane was quickly rinsed with 20 ml of 1× PBS–0.1% Tween and then blocked for 1 h at RT in 20 ml of 1× PBS supplemented with 2% fat-free milk powder. Collagen (bovine collagen type I; Sigma) was bound by incubation for 1 h at RT in 20 ml of 1× PBS–0.1% Tween containing 2% milk powder and 10 μg ml⁻¹ collagen. After two washes (10 min in 20 ml of 1× PBS–0.1% Tween), the primary antibody was added (monoclonal ColE1 [Sigma], diluted 1:2,000 in 20 ml of 1× PBS–2% milk powder) and incubated for 1 h at RT or overnight at 4°C. After two washes (10 min each in 20 ml of 1× PBS), HRP-conjugated secondary (anti-mouse) antibody (Santa Cruz Biotechnology) was added at a dilution of 1:5,000 in 20 ml of 1× PBS–2% milk powder. The secondary antibody was detected (see above).

MS.

To identify the protein in bands cut from SDS-PAGE, gel pieces were destained at 37°C by several washes in 25% and 50% acetonitrile (ACN), reduced at 56°C for 45 min in 50 mM dithiothreitol (DTT), and alkylated for 2 h at RT in the dark using 55 mM iodoacetamide. Residual reagents were washed out with 50% acetonitrile in 25 mM ammonium bicarbonate buffer (NH₄HCO₃). Gel pieces were dehydrated in 100% ACN and dried in a SpeedVac for 15 min. The gel pieces were reswollen with 15 μl of trypsin solution (10 ng/μl in 25 mM NH₄HCO₃, pH 8.0) for 15 min, and after that 20 μl of 25 mM NH₄HCO₃ was added. Digestion was carried out at 37°C overnight. Peptides were extracted by sonication and dried with 100% ACN. The extracts were evaporated to dryness and resuspended in 2% ACN with 0.05% trifluoroacetic acid (TFA). Peptides were analyzed using an UltiMate 3000RS LCnanoSystem (Dionex) coupled with a MicrOTOF-Q II mass spectrometer (MS; Bruker) using an Apollo Source ESI nano-sprayer equipped with a low-flow nebulizer.

The peptides were injected on a C₁₈ precolumn (Acclaim PepMap Nano Trap column) using 2% ACN with 0.05% TFA as a mobile phase and further separated on a 15-cm by 75-μm reverse-phase (RP) column (Acclaim PepMap Nano Series column; 100-μm particle size; 100-Å pore size) using a 2 to 40% ACN gradient in 0.05% TFA for 60 min. The MS was operated in standard data-dependent acquisition (DDA) tandem MS (MS/MS) mode with fragmentation of the most abundant precursor ions.

Mascot Generic format (.mgf) peak lists were generated using DataAnalysis, version 4.0, software and further searched against the NCBI database or a custom database containing chimeric protein sequences and common contaminants (adapted from the common Repository of Adventitious Proteins [cRAP] database [http://www.thegpm.org/crap/index.html]) using our in-house Mascot server (version 3.0; Matrix Science, London, United Kingdom).

Because of the high sequence similarity of the proteins of interest as well as of their chimeric forms, the identification of species was made on the basis of unique peptides. The number of identified MS/MS spectra of unique peptides can also be used for estimation of the protein amount, enabling semiquantitative analysis. Spectral count normalization was applied to the estimation of relative protein levels or their constituents (passenger or β-barrel) in the studied samples. Normalized spectral abundance factors (NSAFs) were used to compare the amount of the particular protein (k) between samples:

$${\left(\mathrm{NSAF}\right)}_k=\frac{{\left(\mathrm{SpC}/\mathrm{Length}\right)}_k}{{\displaystyle \sum _{i=1}^N{\left(\mathrm{SpC}/\mathrm{Length}\right)}_i}}$$

where the total number of tandem MS spectra counts (SpC) matched to protein k was divided by the length of protein k; this value was then divided by the sum of the quotients SpC/length for all N proteins occurring in the sample (10), where i is the ith element of the sum. NSAF values range from 0 to 1; a value closer 1 indicates a higher protein level.

Maltose-binding protein (MBP) detection. (i) Preparation of minimal medium supplemented with maltose.

Minimal medium supplemented with maltose was prepared according to the protocol by Elbing and Brent (7). Briefly, for 1 liter of a 5× concentration of medium, 30 g of Na₂HPO₄, 15 g of KH₂PO₄, 5 g of NH₄Cl, 2.5 g of NaCl, and 15 mg of CaCl₂ were resuspended in MilliQ water. Then, the medium was autoclaved and diluted to a 1× concentration, and the following were added: 1 ml of 1 M MgSO₄, 10 ml of 20% maltose, 4 ml (25 mg/ml) of methionine, and 10 ml of (10 mg/ml) amino acids mix.

(ii) Protein expression in minimal medium supplemented with maltose.

BL21(DE3) Omp8 strains containing our constructs were grown overnight in minimal medium supplemented with maltose and 100 μg/ml ampicillin in 5-ml cultures at 30°C (36). The following morning, the OD was measured, and the bacteria were diluted to an OD of ∼0.4 in 40 ml of fresh minimal medium supplemented with 20% maltose and 100 μg/ml ampicillin and allowed to grow at 30°C to mid-log phase (OD₆₀₀ of ∼0.5). Protein production was then induced by the addition of 0.2 mM IPTG. After a further 2 h of incubation at 30°C, we diluted cells to an OD₆₀₀ of 1 in a volume of 40 ml, and protein expression was stopped with the addition of 25 μg/ml of chloramphenicol. The samples were split in two, proteinase K was added to a final concentration of 0.01 mg/ml to one of them (17), and the samples were incubated for 15 min. Proteolysis was stopped by addition of phenylmethylsulfonyl fluoride (PMSF) to a final concentration of 0.2 mM. PMSF was added to the control samples as well. All samples were centrifuged for 10 min at 4,000 × g and resuspended in 2 ml of cold 10 mM HEPES, pH 7.2. The cells were broken down by sonication twice for 30 s each time (MS73 probe [Bandelin Sonoplus]; 0.5-s duty cycle). Samples were taken, 2× nonreducing SDS-PAGE sample buffer was added, and SDS-PAGE and Western blot transfer were performed.

(iii) Detection of maltose binding protein.

After a blotting step, the membrane was blocked for 1 h at RT in 20 ml of 1× PBS supplemented with 5% fat-free milk powder. Then, the membrane was washed three times for 5 min in 20 ml of 1× PBS–0.05% Tween, and anti-MBP (Sigma-Aldrich) was bound by incubating the membrane for 2 h at RT in 20 ml of 1× PBS–1% BSA. The membrane was then washed three times for 5 min each time in 20 ml of 1× PBS–0.05% Tween. After this, we incubated the membranes for 1 h at RT with HRP-conjugated secondary (anti-mouse) antibody (Santa Cruz Biotechnology) at a dilution of 1:5,000 in 20 ml of 1× PBS–0.05% Tween. The secondary antibody was detected (see above).

RESULTS

Heterotrimers can be formed among closely related/highly similar TAAs.

Our main goal has been to understand the transport mechanism of TAAs. We therefore first investigated whether heterotrimers could form among members of the Eib group of TAAs. We cloned Eib proteins (Table 1) into the pETDuet-S vector, introduced them into the E. coli BL21 Omp8 strain, isolated OM fractions containing the expressed proteins, and separated them using nondenaturing SDS-PAGE so that we could see trimeric TAAs (Fig. 2A). In addition to the expected homotrimers (EibA, 121.5 kDa; EibC, 156 kDa; EibD, 156.9 kDa), heterotrimers formed between both EibA and EibD and between EibC and EibD. The expression levels of the EibA and EibC homotrimers were equal, whereas EibD was expressed at a lower level. For the EibAD heterotrimer (from pETDuet-S encoding both EibA and EibD), we observed the following four bands, in order of size (from top to bottom on Fig. 2A): EibD₃ (hardly visible), EibA/EibD₂, EibA₂/EibD, and EibA₃. EibD₃ is hardly visible because EibD expresses at a lower level than EibA, as also for homotrimers. Most of the EibD is, then, in the heterotrimers, so very little of the EibD homotrimer is present. The relative molecular masses of EibC and EibD differ by only 0.9 kDa, making it impossible to distinguish them on a gel; as a result, the EibCD heterotrimers (from pETDuet-S-encoded EibC and EibD) ran as a broad band (Fig. 2A). This presumably meant that the band contained both homo- and heterotrimers.

Fig 2.

Eib proteins form heterotrimers. Expression and outer membrane purification show formation of heterotrimers between EibA and EibD and between EibC and EibD. As a negative control, OM proteins purified from cultures transformed with empty pETDuet-S vector were used (marked Φ), and, as positive controls, strains expressing only EibA, EibC, or EibD were used. M, molecular mass marker (kDa). (A) Coomassie-stained SDS-PAGE. (B) Western blots with StrepTactin that binds EibA and EibC.(C) Ni-NTA conjugates that bind EibD. For better comparison of gel and blot results, two figure sets were composed for heterotrimers EibAD (D) and heterotrimers EibCD (E). In panels D and E, the letters above the lanes (a to i) show which lanes from the original gel and blots shown in panels A to C were used; in panel D numbers I to IV mark heterotrimers, as follows: IV, EibA₃; III, EibA₂/EibD; II, EibA/EibD₂; I, EibD₃. Surface expression of heterotrimers was confirmed by proteinase K assay (F). Arrows mark bands corresponding to full-length proteins, and arrowheads mark bands corresponding to translocation domains of the proteins. −, untreated samples; +,samples treated with proteinase K. (G) A functional assay with Fc IgG binding was also made. For further confirmation of our interpretation, bands marked with numbers 1 to 7 in panel A were cut from the gel and sent for mass spectrometry measurements (Table 2).

We used Western blotting to make sure our interpretation was correct. EibA and EibC have a StrepII tag and can be detected with StrepTactin (Fig. 2B), while EibD has a His tag and can be detected using a Ni-NTA conjugate (Fig. 2C). In the EibAD experiment (Fig. 2D), StrepTactin bound well to bands III and IV but bound only slightly to band II, while Ni-NTA bound only bands II and III. In other words, EibA₃ and EibA₂/EibD bound StrepTactin well, while EibA/EibD₂ bound it poorly, as expected with the StrepII tag on EibA. Conversely, only EibA₂/EibD and EibA/EibD₂ bound the Ni-NTA conjugate. (EibD₃ was present in such small amounts that binding was not observed.) As expected, the unresolved EibCD band (Fig. 2A) bound both StrepTactin (Fig. 2B) and Ni-NTA (Fig. 2C), consistent with the presence of heterotrimers (Fig. 2E).

We confirmed these results by excising the bands from the gel and examining them by mass spectrometry (Table 2). This not only confirmed the formation of heterotrimers but also showed the ratio of monomers within one band. To compare the amounts of monomers in one band, the NSAF was calculated (see Materials and Methods). For samples 4, 5, and 6 we can see that abundance of EibA increases, with the amount in sample 4 less than that in sample 5, which is less than the amount in sample 6, and that there is a concomitant decrease of EibD abundance. This is consistent with the data from SDS-PAGE (Fig. 2A) and Western blotting (Fig. 2B and C). The expected NSAF ratio for EibA/EibD₂ (sample 4) is 1:2, and for EibA₂/EibD (sample 5) it is 2:1. The observed NSAFs are 1:1.5 and 1.8:1, respectively. The results are consistent, given that NSAFs are semiquantitative. Moreover, in samples 1, 2, and 6, the StrepII tag was one of detected peptides.

Table 2.

Mass spectrometry measurements of Eib homo- and heterotrimers^a

Sample no.	No. of unique peptides	No. of sequences for unique peptides	No. of sequenced peptides	Total no. of MS spectra (SpC)	Protein	NSAF
1	16	166	19b	172	EibA	1.00
2	8	33	21b	210	EibC	1.00
3	10	57	45	119	EibD	1.00
4	8	32	15	56	EibD	0.59
4	7	20	9	29	EibA	0.41
5	9	31	14	56	EibD	0.35
5	15	70	16	80	EibA	0.65
6	18	215	18b	224	EibA	1.00
7	9	180	18	240	EibC	0.75
7	9	29	19	80	EibD	0.25

^aBands cut for mass spectrometry measurements are marked with numbers 1 to 7 on Fig. 2A.

^bThe StrepTactin sequence (QWSHPQFEK) is present in the sequenced peptide.

The data above, however, do not prove that the heterotrimeric passenger domain is on the cell surface and generates a protein with the correct topology. To prove this, we used a proteinase K assay on Eib-expressing cells. Proteinase K treatment of whole cells reduced the sizes of both homotrimeric and heterotrimeric Eibs (Fig. 2F). The relative molecular masses of EibC, EibD, and EibCD all decreased in the same way, as expected if the entire passenger domain is cleaved off. In the absence of proteinase K treatment, there was a significant amount of cleaved protein (i.e., barrel alone) in the EibCD heterotrimer lane and just a small amount in EibC lane (Fig. 2F). This again indicates that the EibCD coexpression generates genuine heterotrimers, which are then less stable, rather than a stoichiometric mixture of homotrimers.

Finally, we tested whether the heterotrimers are able to bind IgG Fc nonimmunologically (23). We performed standard Western blotting with Fc(IgG)-HRP and developed the blot with the substrate for HRP (Fig. 2G). We could observe strong binding of Fc to homotrimers but weaker binding to EibA₂/EibD and to the EibCD mixture. We did not observe binding to EibA/EibD₂, presumably due to the low level of protein expression or the misfolding of the heterotrimeric passenger domain.

Heterotrimers cannot be formed by distantly related TAAs.

For the full-length sequence, there was 49.5% identity between EibA and EibD, 84.6% between EibC and EibD, and 48.3% between EibA and EibC (data not shown). Alignment of translocation domains of EibA with EibD and of EibA with EibC showed 97.9% identity, while EibC and EibD are identical in this region (see Fig. S1 in the supplemental material). Since the highly conserved Eibs can form heterotrimers, we decided to test if this is also possible for more distantly related TAAs. We therefore coexpressed YadA and EibA, where the full-length proteins are 32.5% identical (data not shown) and the translocation domains are 42.9% identical (see Fig. S1). After induction, purification of OM proteins, and SDS-PAGE, we observed YadA alone, EibA alone, or no protein at all (Fig. 3). This was true for 10 different colonies (data not shown). To make sure that this was not due to degradation of proteins that were stuck in the outer membrane or the periplasm, unable to pass to the outside of the cell, we transformed the plasmid into a ΔdegP strain, a BL21(DE3) Omp8 derivative (12) which lacks the periplasmic protease DegP that clears aggregated or denatured proteins from the periplasmic space in E. coli (28). Again, cells were induced with IPTG, and the OM proteins were purified and separated by SDS-PAGE. The results were the same as for the BL21 Omp8 strain (data not shown).

Fig 3.

YadA and EibA do not heterotrimerize. Based on coexpression in pETDuet-S vector of distinct TAAs, EibA and YadA showed no heterotrimer formation. The figure shows a Coomassie-stained SDS-PAGE gel after OM protein purification from three cultures with different protein expression levels (EibA/YadA lanes 1, 2, and 3). As a negative control OM proteins purified from cultures transformed with empty vector pETDuet-S (marked Φ) were used, and, as a positive control, homotrimers of EibA and YadA were used. M, molecular mass marker (kDa). Protein expression was done at 37°C, where YadA expression is higher than expression of EibA.

The results were puzzling as we had of course sequenced the plasmids before transforming them into the E. coli expression strains. These transformants did initially grow very slowly on the agar-ampicillin plates after transformation, producing only very small colonies after overnight growth. We therefore purified plasmids from four liquid cultures after induction (two expressing YadA and two expressing nothing at all) and resequenced them. When there was no EibA expression, a point mutation (35 Leu to stop) had occurred at the beginning of the eibA open reading frame. Plasmids from two cultures that expressed neither protein gave different and very short fragments after sequencing, indicating a large deletion in both the yadA and eibA genes. This was also true for two other transformants that expressed neither protein. Coexpression of YadA and EibA seems lethal for the cells.

Heterotrimerization depends on the translocation domain.

As YadA and EibA, with low percentage identity in the translocation domain, do not form heterotrimers while the almost identical EibA and EibD do, we decided to test if changing the translocation domain in YadA to that of EibA and vice versa would yield viable heterotrimers. We expected that the translocation domain would be more important for trimerization and transport because earlier results (5) indicated that defects in passenger domain trimerization do not prevent TAA expression.

We created passenger-translocation chimeric TAAs EibA-YadA and YadA-EibA, where the translocation domains contain the β-barrels and linker helices. These were cloned with the TAA corresponding to the translocation domain into the pETDuet-S vector (Table 1). Constructs were expressed, OM proteins were purified, and the proteins were separated by nondenaturing SDS-PAGE so that we could visualize homo- and heterotrimers (Fig. 4A). The majority of EibA proteins remained as trimers (121.5 kDa), and only a small amount dissociated into monomers (40.5 kDa) (Fig. 4A, lane 1). YadA, on the other hand, migrated as a mixture with trimeric (138.3 kDa), dimeric (92.2 kDa), and monomeric (46.1 kDa) proteins all present in similar amounts (Fig. 4A, lane 2). This is consistent with previous results (47) that YadA trimers dissociate into dimers and monomers under SDS-PAGE treatment, but the results do not imply that this occurs in the membrane. In addition, the control expression of the EibA-YadA (trimer of 119.4 kDa) and the YadA-EibA (trimer of 140.4) chimeras also produced some levels of trimers at the expected relative molecular masses (Fig. 4A, lanes 3 and 4). Interestingly, the EibA-YadA chimera migrated primarily as a monomer (39.8 kDa), while only a small amount of YadA-EibA migrated as a monomer (46.8 kDa). The lack of stability of YadA under these experimental conditions resides in the barrel.

Fig 4.

The translocation domain is responsible for trimerization and translocation. Coexpression and OM purification of chimeric proteins with EibA or YadA prove sufficiency of β-barrel for trimerization and translocation of passenger domain. For all panels, lanes are as follows: lane Φ, control sample OM proteins from expression of empty vector pETDuet-S; lane 1, EibA; lane 2, YadA; lane 3, EibA-YadA; lane 4, YadA-EibA; lanes 5 and 6, investigated samples EibA/YadA-EibA and YadA/EibA-YadA, respectively, that form heterotrimers; lane M molecular mass marker (kDa). (A) Coomassie-stained SDS-PAGE gel. Western blotting with StrepTactin, which binds EibA (B), and with Ni-NTA conjugate, which binds YadA (C) showed heterotrimer formation. (D) Surface expression of heterotrimers was confirmed with proteinase K assay. Arrows mark bands corresponding to full-length proteins, and arrowheads mark bands corresponding to translocation domains of the proteins.−, untreated samples; +, samples treated with proteinase K. Functional assays with Fc IgG binding (E) and with collagen (F) were performed showing that heterotrimers still posses the activity of comprised proteins. Finally, for further confirmation of our interpretation, bands marked with numbers 1 to 13 in panel A were cut from the gel and sent for mass spectrometry measurements (Table 3).

The SDS-PAGE experiment also showed that the expected heterotrimers were created. For the EibA/YadA-EibA construct we obtained three bands, corresponding to EibA/(YadA-EibA)₂, EibA₂/YadA-EibA, and EibA₃ (Fig. 4A, lane 5). For the YadA/EibA-YadA construct, we obtained YadA₂/EibA-YadA, YadA/(EibA-YadA)₂, and (EibA-YadA)₃ (Fig. 4A, lane 6). In addition, we also observed monomeric YadA-EibA, which is not surprising based on our YadA and EibA-YadA control experiments (above). Differences in expression levels among heterotrimers within one construct presumably result from the differential barrel stability of each of the possible species, and, as for the control homotrimers, heterotrimers containing more EibA are more stable.

To confirm this interpretation, we used Western blotting (with StrepTactin and Ni-NTA conjugate) as the two proteins were differentially tagged (EibA with a StrepII tag and YadA with a His tag). As expected, StrepTactin bound to both monomeric and trimeric EibA and to homotrimers and monomers of EibA-YadA (Fig. 4B, lanes 1 and 3). For heterotrimers (Fig. 4B, lanes 5 and 6) binding occurred only for EibA₂/YadA-EibA and YadA/(EibA-YadA)₂ and not for EibA/(YadA-EibA)₂ and YadA₂/EibA-YadA. Ni-NTA conjugate bound to all YadA and YadA-EibA forms (Fig. 4C) but only to the monomeric form of YadA when it was coexpressed with EibA-YadA (Fig. 4C, lane 6). We could not observe binding to the heterotrimeric proteins. One possible reason for this is steric hindrance, which could be created due to mismatches in heterotrimeric passenger domains during folding.

We further confirmed these results and our interpretation by excising the bands from the gel and examining them by mass spectrometry (Table 3 and Fig. 5). For samples 1, 2, 3, 8, and 13, the StrepII tag was detected, and for sample 4 the His tag and the linker sequence that contains amino acids of the YadA passenger domain and EibA translocation domain were detected. For these results, comparison of the NSAFs is difficult as the β-barrels of proteins ionize poorly and so were not detected by MS. However, the EibA passenger abundance increases in the heterotrimeric samples such that the amount in sample 6 is less than that in sample 7 which is less than that in sample 8, and the YadA passenger abundance decreases from sample 9 to sample 10 and then again to sample 11, which is in strong agreement with our interpretation.

Table 3.

Mass spectrometry measurements of YadA and EibA heterotrimers^a

Sample no.	No. of unique peptides	No. of sequences for unique peptides	No. of sequenced peptides	Total no. of MS spectra (SpC)	Protein (part of protein)	NSAF
1	21	172	21b	172	EibA (passenger)	0.60
1	6	28	6	28	YadA (barrel)	0.40
2	22	285	22b	285	EibA (passenger)	0.56
2	7	54	7	54	YadA (barrel)	0.46
3	26	197	26b	197	EibA (passenger)	0.48
3	6	52	6	52	YadA (barrel)	0.52
4	22	229	22c	229	YadA (passenger)	0.95
4	0	0	1	2	EibA (barrel)	0.03
4	1	3	1	3	EibA (passenger)	0.02d
5	15	117	15	117	YadA (passenger)	0.67
5	1	6	2	15	EibA (barrel)	0.33
6	6	40	6	40	EibA (passenger)	0.37
6	10	86	10	86	YadA (passenger)	0.63
7	14	97	14	97	EibA (passenger)	0.54
7	14	106	14	106	YadA (passenger)	0.46
8	19	178	19b	178	EibA (passenger)	1.00
9	9	61	9	61	YadA (passenger)	0.45
9	7	54	7	54	EibA (passenger)	0.51
9	0	0	1	1	YadA (barrel)	0.04
10	13	83	13	83	YadA (passenger)	0.30
10	11	83	11	83	EibA (passenger)	0.38
10	0	0	3	17	YadA (barrel)	0.32
11	1	1	1	1	YadA (passenger)	0.004d
11	14	111	14	111	EibA (passenger)	0.59
11	0	0	4	18	YadA (barrel)	0.40
12	1	3	1	3	YadA (passenger)	0.01d
12	21	174	21	174	EibA (passenger)	0.49
12	0	0	6	42	YadA (barrel)	0.50
13	2	2	2	2	YadA (passenger)	0.004d
13	20	168	20b	168	EibA (passenger)	0.43
13	0	0	6	52	YadA (barrel)	0.56

^aBands cut for MS measurements are marked with numbers 1 to 13 on Fig. 4A.

^bThe StrepTactin sequence (QWSHPQFEK) is present.

^cThe His tag sequence (MGSSHHHHHHSQDPDDYDGIPNLTAVQISPNADPALGLEYPVRPPVPGAGGLNASAK) and the linker sequence containing the C terminus of the YadA passenger and the N terminus of the EibA barrel are present.

^dThe abundance of the YadA passenger domain in samples 11, 12, and 13 is insignificant relative to the peak area (Fig. 5A and B) and appears to be due to sample contamination that occurred when the bands were excised. It is then a baseline for other measurements. The EibA passenger domain in sample 4 is also due to contamination, as this sample was from YadA-EibA, and so no EibA passenger domain could be present.

Fig 5.

The comparison of precursor MS spectra: baseline creation. Spectra of two sequences, TTLETAEEHANSVAR (A) and SSSVLGIANNYTDSK (B), the only unique sequences found for the YadA passenger domain, are shown. Spectra for samples containing full-length YadA protein (bold dashed line, sample 9; bold solid line, sample 10) are compared with samples that should not contain a YadA passenger domain unless contaminated (unbolded line, samples 11, 12, and 13).

Proteinase K treatment on intact cells followed by inactivation of the enzyme and SDS-PAGE showed that all of the chimeras were surface exposed (Fig. 4D); for the majority of the constructs [EibA, EibA-YadA, EibA/(YadA-EibA) and YadA/(EibA-YadA)], the band corresponding to the trimeric TAA (Fig. 4D, arrow) disappears and is replaced by a band corresponding to the translocation domain trimer. For the YadA homotrimer and the YadA-EibA chimeric homotrimer, there was a significant decrease in the amount of full-length trimer, and a band corresponding to the size of the translocation domain appeared. Moreover, the disappearance of YadA dimers and monomers and the formation of just trimeric YadA translocation domain bands further shows that those forms are created during SDS-PAGE; the results imply that, in SDS, the full-length protein is less stable than the YadA translocation domain alone. To prove the integrity of the OM and show that disappearance of the bands corresponding to full-length proteins does not result from leaking of proteinase K into the periplasm, we used Western blotting to detect MBP (Fig. 6). It is clear that the outer membrane is intact as the signals from treated (Fig. 6, + lanes) and untreated (Fig. 6, − lanes) cells are the same. In addition, we observed nonspecific binding of the secondary antibody by the EibA passenger domain (EibA, EibA-YadA, EibA/YadA-EibA, and YadA/EibA-YadA) and its digestion by proteinase K, as shown in Fig. 4. This proves that constructs that we expressed are expressed on the cell surface.

Fig 6.

The outer membrane is intact. Expression of chimeric proteins with EibA or YadA in minimal medium supplemented with maltose followed by proteinase K treatment and MBP detection proves integrity of OM. MBP levels are shown for the empty pETDuet-S vector (Φ; positive control for MBP level),wild-type proteins EibA and YadA, chimeric EibA-YadA and YadA-EibA, and coexpressed EibA/YadA-EibA (E/Y-E) and YadA/EibA-YadA (Y/E-Y). Lane M, molecular mass marker (kDa); −, untreated sample; + sample treated with proteinase K. The MBP band is boxed. Results were similar for the EibA, -C, and -D heterotrimers (data not shown).

Finally, we showed that the chimeric heterotrimers possess the activity of both passenger domains. Membranes were probed with IgG Fc or collagen, followed by an anti-collagen antibody. IgG Fc bound the EibA passenger domain in both monomeric and trimeric forms (Fig. 4E). However, as with binding the StrepTactin, not all trimers containing the EibA passenger domain bound Fc. Fc bound to EibA₂/YadA-EibA and EibA₃ (Fig. 4E, lane 5), the YadA/(EibA-YadA)₂ heterotrimer, (EibA-YadA)₃, and apparently to the EibA-YadA monomer (Fig. 4E, lane 6). Collagen bound to trimeric and, surprisingly, to dimeric and monomeric forms of YadA (Fig. 4F, lane 2) and to trimers and monomers of the YadA-EibA chimera (Fig. 4F, lane 4). Binding to trimeric YadA and YadA-EibA is much more efficient than to other forms of YadA and monomeric YadA-EibA, as expected. Binding to dimers and monomers of YadA can be explained by refolding of YadA on the nitrocellulose membrane (8). Lack of binding to heterotrimers (Fig. 4F, lanes 5 and 6) is according to our predictions because there is no band containing a trimeric form of the YadA passenger domain. In the collagen blot the nonspecific binding to the YadA/(EibA-YadA) (lane 6) is most probably due to nonimmune interactions between EibA and the antibodies. This could be seen by comparing the strength of the signals in lanes 2 and 4 with that in lane 6.

DISCUSSION

Folding and autotransport: implications for mechanism.

If the signal for transport of the TAA passenger domain to the cell surface is not in the passenger domain, it should logically be in the translocation (barrel plus linker) domain. This is consistent with studies on the structure and pore opening of the monomeric autotransporter NalP (30), where the helical linker region was shown to move in and out of the barrel domain. In YadA, deletions of different regions of the passenger domain did not disturb surface expression (35). Translocation occurred even when one or more of the head, neck, or stalk domains were deleted. The linker and β-barrel domain were, however, always required. Similarly, Cotter et al. (5) showed that null passenger domains still led to folded β-barrels in the outer membrane. In addition, we have recently shown that the EibD passenger domain by itself undergoes a trimer-dimer equilibrium during gel filtration (24). On the other hand, Ieva and Bernstein (16) showed that, during translocation by the monomeric autotransporters, BamA interacts not just with the translocation domain but also with the passenger domain. However, they still support the hairpin model as stalling of autotransportation resulted in only the C terminus of the passenger domain being degraded by proteinase K.

Studying the heterotrimeric barrels would, we thought, give insight into the rules of barrel folding and transport. They formed heterotrimers if the barrels were of the Eib group (identity > 97%) (Fig. 2), but YadA and EibA (barrel sequence identity of 43%) did not form hetero-barrels at all. To our surprise the combination was clearly toxic to the cell, with either one or both genes eliminated shortly after transformation (Fig. 3). YadA and Hia were reported in a previous study to be coexpressed (5), but the authors used two plasmids, which may have led to one of the two plasmids being eliminated stochastically, and the observed expression of both proteins could be the result of a heterologous population of cells within the cultures, some expressing only Hia and some only YadA. Alternatively, YadA and Hia maybe divergent enough (21.8% identity of translocation domain) that heterotrimerization is not even attempted. On the other hand, as long as the translocation domain was not heterologous (i.e., for YadA, either YadA or YadA/EibA-YadA; for EibA, either EibA or EibA/YadA-EibA), a heterotrimeric protein formed, and the now heterotrimeric passenger domain was transported to the cell surface (Fig. 4D).

Consequently, we suggest that trimerization of the translocation domain, which seems to occur during or prior to membrane insertion, is required for translocation of the passenger domain to the cell surface. Passenger trimerization, on the other hand, is not required for membrane insertion and so presumably happens later. What, then, is the mechanism of passenger trimerization and transport? We start from the assumption that the three domains must be transported together because individually they do not form folded entities, and so the protein folding would not drive transport in a ratchet-like manner. Given this, one possible mechanism is through the translocation domain, with three passenger domains passing through the barrel domain as hairpins at the same time (Fig. 1C). This is based on results of Grosskinsky and coworkers (12), who showed that the conserved Gly389 (YadA numbering) in the center of the TAA barrel is required for folding but not structure. Even the G389A mutation let to reduced expression (12). Furthermore, our results indicate that folding and transport are intrinsic to the translocation domain; the YadA/EibA heterotrimer formed only when the barrels were the same (Fig. 4D). On the other hand, space limitations in the β-barrel lumen indicate that the hairpin may form the other way: one half of the hairpin would be inside the β-barrel lumen while the other half would be between the β-barrel and the Bam complex (Fig. 1D), as suggested by Ieva and Bernstein (16).

Overall, we prefer the internal hairpin model (Fig. 1C) for the following reasons: (i) this model does not require BamA for translocation of the passenger domain but explains its role in barrel assembly and makes its role the same for all β-barrels inserting into the OM (13); (ii) the model leads to the greatest level of steric crowding around Gly389 during translocation; and (iii) the model explains the requirement for both linker and β-barrel in translocation.

Biological implications.

Are the mixed barrels we see in vitro (above) relevant in vivo? In the highly homologous Eib family of Ig Fc-binding proteins, such mixed trimers may be physiologically relevant. We demonstrated that EibAD and EibCD heterotrimers occur. As EibC and EibD are almost identical (97% between EibA and EibD), it is highly likely that EibAC heterotrimers also occur. This finding is very interesting as multiple Eib genes occur naturally in E. coli strains (37). Their natural coexpression can result in the binding of different subsets of Igs as different Eibs bind with various affinities or not at all to different Ig subtypes (23, 37, 38). There are two possibilities: either bacteria have a mechanism that prevents heterotrimerization, or heterotrimers form and are at least not disadvantageous. If it is the former, one possibility would be transcription regulation so that only one eib gene is transcribed at a time. The latter model seems more likely because Eibs naturally form heterotrimers, and this may confer a selection advantage. First, it would increase the levels of expression of the Eibs by a gene dose effect, and, second, the heterotrimers, which still bind their ligands (Fig. 2G and 4E and F), would provide various affinities and specificities in comparison with homotrimers. This could be advantageous during infection as a way of increasing surface heterogeneity and thus immune evasion.

In the YadA-EibA coexpression studies, coexpression of distant homologues appears to be deleterious. We suggest that, in such cases, the genes will be tightly regulated to ensure that concurrent expression does not occur. This suggestion is consistent with the existence of allelic TAAs in Haemophilus influenzae, Hia and Hsf (1), that are very different in length (1,096 and 2,413 residues, respectively) but have very similar translocation domains. Even though their β-barrels would allow heterotrimerization, it is highly unlikely that an active passenger domain could form due to the 700-residue difference in length. Allelic expression is certainly consistent with our model. Elucidating the folding rules for TAAs and what happens in vivo is clearly an important goal.