Species-Specificity of the BamA Component of the Bacterial Outer Membrane Protein-Assembly Machinery

The BamA protein is the key component of the Bam complex, the assembly machinery for outer membrane proteins (OMP) in gram-negative bacteria. We previously demonstrated that BamA recognizes its OMP substrates in a species-specific manner in vitro. In this work, we further studied species specificity in vivo by testing the functioning of BamA homologs of the proteobacteria Neisseria meningitidis, Neisseria gonorrhoeae, Bordetella pertussis, Burkholderia mallei, and Escherichia coli in E. coli and in N. meningitidis. We found that no BamA functioned in another species than the authentic one, except for N. gonorrhoeae BamA, which fully complemented a N. meningitidis bamA mutant. E. coli BamA was not assembled into the N. meningitidis outer membrane. In contrast, the N. meningitidis BamA protein was assembled into the outer membrane of E. coli to a significant extent and also associated with BamD, an essential accessory lipoprotein of the Bam complex.Various chimeras comprising swapped N-terminal periplasmic and C-terminal membrane-embedded domains of N. meningitidis and E. coli BamA proteins were also not functional in either host, although some of them were inserted in the OM suggesting that the two domains of BamA need to be compatible in order to function. Furthermore, conformational analysis of chimeric proteins provided evidence for a 16-stranded β-barrel conformation of the membrane-embedded domain of BamA.


Introduction
Gram-negative bacteria are characterized by a cell envelope consisting of an inner and an outer membrane (OM), which are separated by the peptidoglycan-containing periplasm.While integral inner membrane proteins are α-helical, all but two known integral OM proteins (OMPs) are β-barrels [1][2][3].Only recently, the mechanisms of the assembly of OMPs into the OM have started to come to light.Previously, we showed that the Omp85 protein is an essential part of the OMP assembly machinery in Neisseria meningitidis; depletion of this essential protein in a conditional mutant strain resulted in the accumulation of unassembled forms of all OMPs analyzed [4].Later, a similar function was shown for the Omp85 homologs in Escherichia coli [5][6][7], Pseudomonas aeruginosa [8] and Borrelia burgdorferi [9].Omp85 homologs have been renamed BamA, for β-barrel assembly machine component A.
BamA was predicted to consist of an N-terminal part localized in the periplasm and a membrane-embedded β-barrel in its C-terminal part [4].In the predicted periplasmic part, five polypeptide-transport-associated (POTRA) domains are located [15].Crystallographic and NMR structural studies have shown a similar fold for all POTRA domains: a β-sheet of three β-strands overlaid with a pair of antiparallel α-helices [16][17][18][19].For the β-barrel domain, originally a 12-stranded β-barrel topology was predicted [4], which leaves a 61-amino-acid region between the POTRA and β-barrel domains, which we designated the hinge region [20].However, the crystal structure of the two-partner secretion-system component FhaC of Bordetella pertussis, which shares limited sequence similarity with BamA, showed that this protein consists of two POTRA domains and a 16-stranded β-barrel, suggesting that the BamA β-barrel might also contain 16 β-strands, with four additional βstrands located in the postulated hinge region [21,22].
We have shown that EcBamA interacts with OMPs via the so-called signature sequence present at the C terminus of the substrates [23].This sequence is characterized by a phenylalanine or a tryptophan at the C-terminal position and hydrophobic amino-acid residues at positions -3, -5, -7, and -9 from the C terminus [24].EcBamA interacted in vitro with E. coli-derived C-terminal signature sequences, but not with those found in neisserial OMPs, the defining difference being the nature of the penultimate C-terminal amino acid [23].These results are consistent with the observation that neisserial OMPs are not efficiently assembled in E. coli.This apparently high species-specificity of the Bam machinery seems to be at odds with the observation that the Bam machinery of E. coli can properly assemble a mitochondrial β-barrel OMP [25] and with the finding that bacterial OMPs were properly assembled into the mitochondrial OM of Saccharomyces cerevisiae [26].
To gain further insight into the species-specificity of the functioning of the Bam machinery, we investigated here whether BamA homologs from various Gram-negative bacteria are able to substitute each other in vivo.

Plasmid Constructions
The plasmids and primers used are listed in Tables S1 and  S2, respectively.DNA fragments containing bamA genes were obtained by PCR using genomic DNA from N. meningitidis HB-1 plus primer pair NmF/NmR, N. gonorrhoeae FA1090 DNA plus primer pair NmF/NmR , Burkholderia mallei Bogor DNA (provided by the Central Veterinary Institute, Lelystad, The Netherlands) plus primer pair BmF/BmR, Bordetella pertussis Tohama I DNA (provided by the Netherlands Vaccine Institute, Bilthoven, The Netherlands) plus primer pair BpF/BpR and E. coli DH5α DNA plus primer pair EcF/EcR, and cloned into pCRII-TOPO.We used the sequences of the locus tags NMB0182, NGO1801, BMA1547, BP1427 and b0177, respectively (www.ncbi.nlm.nih.gov/gene), to design primers.The primers carried restriction sites allowing for subcloning of the genes behind the IPTG-inducible promoter on plasmid pFP10-c-lbpA, which is able to replicate in N. meningitidis as well as in E. coli.As a result of the cloning strategy, the original signal sequences were substituted by that of the lactoferrinbinding protein A (LbpA) of N. meningitidis (MNKKHSFPLTLTALAIATAFPSYA).The NdeI-AatII fragments, carrying N. meningitidis, N. gonorrhoeae, and E. coli bamA genes were ligated into NdeI-AatII-digested pFP10-c-lbpA, yielding pFP10-NmbamA, pFP10-NgbamA, and pFP10-EcbamA, respectively.The fragments containing B. pertussis and B. mallei bamA were excised using the NdeI site introduced by the forward primers and the KpnI site of pCRII-TOPO and introduced into NdeI-KpnI-digested pFP10-EcbamA, yielding pFP10-BpbamA and pFP10-BmbamA, respectively.
A construct encoding the C-terminal 317 amino acids of NmBamA fused to the signal sequence of LbpA was created by PCR using primer pair Omp85-F481-NmR and pFP10-NmbamA as template.The resulting PCR product was cloned into pCRII-TOPO and ligated into NdeI-AatII-digested pFP10-c-lbpA, yielding pFP10-481 NmbamA.
The N-and C-terminal domains of NmBamA and EcBamA were swapped resulting in chimeras.Two chimeric bamA genes were designed encoding proteins in which the Nterminal and C-terminal domains of NmBamAand EcBamA were exchanged within a small region of high homology around the start of the predicted 12-stranded β-barrel [4].The construct designated Ec 479 Nm encodes a protein carrying the N-terminal part of EcBamA and the C-terminal part of NmBamA, while construct Nm 480 Ec encodes the reverse chimera.The numbers indicate the most C-terminal amino acid of the BamA domain present in the N-terminal part of the chimera (counting the N-terminal signal sequence).For the Ec 479 Nm construct, two overlapping PCR products were made: one using the EcF and HingeR primers and pFP10-EcbamA as template and the other using the HingeF and NmR primers and pFP10-NmbamA as template.The purified PCR products were mixed, melted and annealed and combined with EcF and NmR as external primers in a second PCR to obtain the complete chimeric genes.For the Nm 480 Ec chimera, also two PCR products were generated: one using NmF and HingeR as primers and pFP10-NmbamA as template, and the other with HingeF and EcR as primers and pFP10-EcbamA as template.The second PCR was performed using purified PCR products and NmF and EcR as external primers.The obtained chimeras were introduced into pCRII-TOPO.The NdeI-AatII fragments were subsequently ligated into pFP10-c-lbpA digested with NdeI and AatII, yielding constructs pFP10-Ec 479 Nm and pFP10-Nm 480 Ec.
Another set of chimeras, designated Ec 423 Nm and Nm 423 Ec, was constructed to produce hybrid proteins where domain exchange occurred at the C terminus of the predicted POTRA5 domain [18] with the N-terminal domain of the chimeras comprising 423 amino-acid residues.To make the Ec 423 Nm construct, a megaprimer was created using primers EcNmF and Omp85R2, and pFP10-NmbamA as the template.The purified megaprimer was combined with the EcF primer and pFP10-EcbamA as the template for a second PCR and the product was cloned into pCRII-TOPO.The chimeric fragment was then introduced into pFP10-NmbamA, using the NdeI site and the KpnI site present in the N. meningitidis bamA gene, yielding pFP10-Ec 423 Nm.For the Nm 423 Ec chimera, megaprimer1 was generated using primer pair NmEcR-Omp85F3 and pFP10-NmbamA as the template.For the second PCR, megaprimer1 was combined with the EcR primer and pFP10-EcbamA as the template.The chimeric fragment obtained was cloned into pCRII-TOPO, yielding pCRII-Nm 423 Ec-a.Megaprimer2 was produced using primers Omp85F4 and Omp85R7 and pFP10-NmbamA as template.Megaprimer2 was then combined with EcR and pCRII-Nm 423 Ec-a as template in a PCR.The resulting PCR product was cloned into pCRII-TOPO, yielding pCRII-Nm 423 Ec-b.The missing 5' part of NmbamAwas added through insertion of a SalI-NotI fragment of pCRII-NmbamA, a precursor of pFP10-NmbamA, into SalI-NotI-restricted pCRII-Nm 423 Ec-b, yielding pCRII-Nm 423 Ec-c.Finally, an NdeI-KpnI fragment of pCRII-Nm 423 Ec-c was ligated into NdeI-KpnI-digested pFP10-EcbamA, yielding pFP10-Nm 423 Ec.Plasmid pRV-His-NmBamA was created by insertion of a NotI-AatII fragment, encoding NmBamA with an additional HHHHHHQDF amino-acid sequence between the signal sequence and the N terminus of the mature BamA protein, from pEN11-His-Omp85 into NotI-AatII restricted pRV2000.All constructs were verified by sequencing.

RT-PCR
Reverse transcription-PCR (RT-PCR) was performed as described [29] except that conventional PCR instead of realtime PCR was performed for amplification of cDNA.Primer couples Q-for-1/Q-rev-Ngo, Q-for-2/Q-rev-Eco, Q-for-3/Q-rev-Bper and Q-for-1/Q-rev-Bmal were used to amplify the bamA cDNAs of N. gonorrhoeae, E. coli, B. pertussis and B. mallei, respectively.

Complementation Assays
To test whether various BamA variants could complement BamA deficiency in E. coli, pFP10-based plasmids containing bamA variants under the control of an IPTG-inducible promoter were introduced into EcBamA-depletion strain UTP BAD ::bamA, which produces EcBamA from an arabinose-inducible promoter.Growth of the resultant strains in LB with indicated supplements was assessed at 37°C or on LB plates at 22°C.To evaluate BamA complementation in N. meningitidis, it was tested whether the chromosomal copy of bamA could be disrupted, when bamA variants were expressed from the pFP10-derived plasmids.To this end, pRV1300 was used as the template to generate a PCR product containing a fragment upstream of N. meningitidis bamA, a kanamycin-resistance cassette and a 3' fragment of bamA, as described [4].The PCR fragment was used to transform HB-1 cells carrying bamA variants on plasmid.Transformations were done in the presence of IPTG and transformants were selected on GC agar plates supplemented with 1 mM IPTG and kanamycin and analyzed by PCR using primer pair G and Omp85R2 that hybridize upstream and in the 3' part of NmbamA.

Urea Extraction
Thirty-μl samples of cell envelope preparations were incubated in 1 ml of 20 mM Tris-HCl, 100 mM glycine (pH 7.6), 6 M urea for 1 h at room temperature while rotating.The insoluble material was separated from the soluble fraction by ultracentrifugation (200,000 x g, 1 h, 4°C).The pellet was dissolved in 30 μl of 2 mM Tris-HCl (pH 7.6), while the solubilized proteins were precipitated from the supernatant with 10% trichloroacetic acid (TCA) and dissolved in 30 μl H 2 O.

Trypsin Treatment
Exponentially growing cells were pelleted and resuspended in phosphate-buffered saline plus 1 mM MgCl 2 and 0.5 mM CaCl 2 to an optical density at 550 nm (OD 550 ) of 2. Trypsin was added to 200-µl portions of this suspension.After incubation for 15 min at room temperature, 1 mM phenylmethanesulfonyl fluoride (PMSF) was added, and the bacteria were collected by centrifugation and boiled in sample buffer for sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).Cell envelopes were treated with 50 μg/ml trypsin at room temperature for the indicated time periods and subsequently boiled in SDS-PAGE sample buffer.

SDS-PAGE and Immunoblotting
Proteins were analyzed by standard denaturing SDS-PAGE as described [14], followed by staining with Coomassie Brilliant Blue or silver [30], or immunoblotting as described [14].EcBamA was detected with a rabbit antiserum raised against the denatured full-length protein [23].Rabbit antisera against the N-terminal (residues 22-464) and C-terminal (residues 455-797) regions of N. meningitidis BamA (α-N-NmBAmA and α-C-NmBamA, respectively) were generously provided by Ralph Judd (University of Montana, USA).The antisera against E. coli BamD and BamB were generous gifts of Naoko Yokota and Hajime Tokuda (University of Tokyo, Japan).A monoclonal antibody (Mab) directed against the POTRA1 domain of NmBamA (α-POTRA1 Nm ) came from GlaxoSmithKline Biologicals (Rixensart, Belgium).

E. coli BamA does not function in N. meningitidis
To determine whether E. coli BamA could substitute for N. meningitidis BamA in vivo, a plasmid containing the EcbamA gene under control of a lac promoter was introduced into N. meningitidis strain HB-1.This resulted in IPTG-dependent expression of EcbamA in HB-1 (Figure 1A).As a control, HB-1 was also transformed with a similar plasmid containing N. meningitidis bamA.Next, we attempted to disrupt the chromosomal NmbamA copy in the transformants.Correct mutants were easily obtained when N. meningitidis bamA was expressed from plasmid.In contrast, PCR analysis showed that the few kanamycin-resistant transformants that were obtained from the cells expressing EcbamA still contained the wild-type copy of NmbamA on the chromosome.Apparently, EcBamA cannot replace N. meningitidis BamA.
The inability of EcBamA to substitute for NmBamA could be due to its failure to become correctly assembled in the meningococcal OM.To test this hypothesis, we analyzed the extractability of EcBamA with urea from these membranes.Correctly inserted OMPs are not usually extracted with urea, as demonstrated for the neisserial porins in Figure 1B (left panel).However, most of EcBamA present in neisserial cell envelopes was extractable with urea (Figure 1B, left panel), whereas, as expected, this was not the case for EcBamA in E. coli cell envelopes (Figure 1B, right panel).Correctly assembled β-barrel domains of OMPs are also usually resistant to proteases.Accordingly, the intact C-terminal β-barrel domain of EcBamA can be obtained by treating E. coli cell envelopes with trypsin [23] (Figure 1C).However, when neisserial cell envelopes containing EcBamA were treated with trypsin, no protected fragment was found (Figure 1C), indicating that no part of EcBamA is properly assembled into the OM.Together, these data demonstrate that EcBamA, for the most part, was not inserted in the neisserial OM and, therefore, cannot be expected to functionally replace NmBamA.

N. meningitidis BamA does not function in E.coli
To assess whether N. meningitidis BamA can substitute for E. coli BamA, we made use of the EcBamA-depletion strain UTP BAD ::bamA in which chromosomal EcbamA expression is under control of an arabinose-inducible promoter resulting in a growth-arrest after prolonged growth in the absence of arabinose (Figure 2A).Introduction of pFP10-NmbamA into this strain did not rescue growth in the absence of arabinose even when NmBamA synthesis was induced with IPTG (Figure 2B), while a control strain containing EcbamA under IPTG control on pFP10-EcbamA continued to grow under these conditions (Figure 2C).The expression of N. meningitidis bamA in response to IPTG induction was confirmed on immunoblots (Figure 3A).Even under conditions of slow growth, i.e. on LB plates at 22°C, NmBamA did not support growth of E. coli (Figure S1).Thus, BamA of N. meningitidis does not function in E. coli.Interestingly, N. meningitidis BamA appeared to be inserted into the E. coli OM since it was not extracted from the membrane fraction with urea just as in neisserial cell envelopes (Figure 3B).
EcBamA is associated with several lipoproteins, i.e., BamB, BamC, BamD and BamE [6,13].A failure to associate with Cell envelopes were extracted with Elugent and the His-tagged NmBamAwas affinity purified with Ni 2+ -NTA beads under native conditions (Figure 3C).A control purification was done on cells not producing the His-tagged NmBamA.Remarkably, substantial amounts of E. coli BamD co-purified with the neisserial BamA (Figure 3C, lane 2), while no BamD was detected in the eluate of the control experiment with cells not expressing the His-tagged NmBamA (Figure 3C, lane 1).In contrast to this essential lipoprotein, the non-essential lipoprotein BamB did not specifically co-purify with NmBamA (Figure 3C).These results enforce the notion that at least a portion of NmBamA produced in E. coli is correctly assembled.

Complementation by Other BamA Homologs
The observation that EcBamA and NmBamA could not functionally replace each other in vivo could be related to the evolutionary distance between E. coli, belonging to the class of γ-proteobacteria, and N. meningitidis, a β-proteobacterium.Therefore, we next investigated whether BamA from other βproteobacteria, i.e. those of B. pertussis, B. mallei and N. gonorrhoeae, could functionally substitute BamA in N. meningitidis.Strain HB-1 was transformed with plasmids containing these bamA genes under lac promoter control.N. gonorrhoeae and B. pertussis BamA protein production could be verified by immunoblotting utilizing the apparent crossreactivity of the α-N-NmBamA and α-EcBamA antisera, respectively (Figure 4A).Expression of the B. mallei bamA gene was demonstrated by RT-PCR (Figure 4B).Next, we tested whether the chromosomal bamA gene could be inactivated while the cells were kept in the presence of IPTG.We only succeeded to obtain correct mutants when the gonococcal bamA was expressed from plasmid.Thus, apparently, the meningococcal BamA protein can be functionally replaced by BamA of the closely related species N. gonorrhoeae, which shows 95% amino-acid sequence identity with the meningococcal one (Figure S2A), but not by those of B. pertussis or B. mallei, which are less similar (Figure S2B).We evaluated also whether the B. pertussis and B. mallei BamA proteins could substitute for E. coli BamA.Neither one allowed for growth of the E.coli BamA-depletion strain in the absence of arabinose and the presence of IPTG at 37°C (Figure 2D,E) or at 22°C (Figure S1) although they were expressed (Figure 4C), demonstrating that they cannot functionally substitute EcBamA.

Functionality of Chimeras of N. meningitidis and E. coli BamA
To determine which part of the BamA protein could be dictating the species specificity of functionality, two sets of chimeras were constructed in which the N-and C-terminal domains of NmBamA and EcBamA are swapped.For the first couple of constructs, designated Ec 479 Nm and Nm 480 Ec, the site of the domain exchange was chosen based on a model predicting a 12-stranded β-barrel in the C-terminal part of BamA [4] (Figure 5A,B).The second set of chimeras, designated Ec 423 Nm and Nm 423 Ec, was designed based on a FhaC-like 16-stranded β-barrel [21]; in this case the domains were swapped directly after the C terminus of POTRA5 (Figure 5A,B).The chimeric genes were cloned behind an IPTGinducible promoter and introduced into N. meningitidis.Synthesis of all hybrid proteins was detected when immunoblots with cell envelopes were probed with appropriate antisera (Figure 6A).However, the chromosomal copy of NmbamA could not be inactivated during expression of any of the chimeras indicating that all four chimeras are non-functional in N. meningitidis.
To test whether the chimeric proteins were assembled in the OM, cell envelopes were treated overnight with trypsin.Such treatment of wild-type membranes yields three bands reactive with an antiserum directed against the C-terminal domain of NmBamA indicated with I, II and III in Figure 6B [14].Treatment of the membranes containing the Ec 423 Nm or Ec 479 Nm chimeras with trypsin also yielded bands I, II and III; this was expected since these membranes also contain native NmBamA encoded by the chromosome.Remarkably, additional distinct fragments of 37 and 30 kDa were detected with the antiserum (indicated with asterisks in Figure 6B) indicating that the C-terminal βbarrel domain of these chimeras was assembled into the OM.In contrast, no trypsin-resistant fragments reactive with the α-EcBamA antiserum were obtained from cell envelopes containing the Nm 423 Ec or the Nm 480 Ec chimera (Figure 6B, right panel) suggesting that these proteins are not assembled in the OM.These observations were confirmed in urea extraction experiments: both Nm 480 Ec and Nm 423 Ec were fully extractable from the N. meningitidis cell envelopes, whereas Ec 423 Nm and Ec 479 Nm did not solubilize in urea (Figure 6C).Thus, only the chimeras containing the C-terminal part of NmBamA appeared to be assembled into the OM of N. meningitidis.
Next, we assessed functionality of the chimeric BamA proteins in E. coli by introducing the plasmids containing the chimeric constructs into the EcBamA-depletion strain.All chimeric proteins were produced after growth of the cells in the presence of arabinose plus IPTG (Figure 7).Extraction of the membrane fraction with urea indicated that the chimeric proteins Ec 423 Nm and Ec 479 Nm were fully inserted into the E. coli OM, whereas the Nm 480 Ec and Nm 423 Ec proteins were each for ~50% extractable and, therefore, only partially inserted (Figure 7).The plasmid-encoded full-length EcBamA was also partially extracted (Figure 7 bottom panel), in contrast to endogenously expressed chromosomal EcBamA (Figure 1B).Apparently, the cells cannot assemble overexpression levels of EcBamA.Despite significant insertion, none of the chimeras appeared to complement the absence of endogenous EcBamA, since the strains stopped growing in the absence of arabinose and presence of IPTG at 37°C (Figure 2F-I) or at 22°C (Figure S1).

Analysis of the NmBamA β-barrel Domain
The simplest model explaining the correct assembly of the Cterminal parts of both the Ec 423 Nm and the Ec 479 Nm chimera into the neisserial OM would be that their membraneembedded domains are fully comprised within the NmBamAderived part.These should then have identical conformations.To test this notion, we treated intact cells with limited amounts of trypsin in order to target only the surface-exposed part of the BamA variants.Trypsin treatment of cells of the parent strain HB-1 or HB-1 cells expressing the Ec 423 Nm or Ec 479 Nm chimera did not affect chromosomally encoded wild-type NmBamA as deduced from similar reactivity with Mab α-POTRA1 Nm , which specifically recognizes POTRA1 of N. meningitidis BamA and, therefore, only reacts with endogenous NmBamA (Figure 8A, left panels).Apparently, no readily accessible trypsin-cleavage sites are exposed at the cell surface in native NmBamA.Next, we probed the samples with an antiserum directed against EcBamA to detect specifically the chimeric proteins.Remarkably, an approximately 7-kDa fragment was cleaved off from the majority of the Ec 479 Nm protein whereas little or no cleavage was observed for the Ec 423 Nm chimera (Figure 8A, right panels).Thus, the conformation of the membrane domain of the Ec 479 Nm chimera appears affected in such a way that one of its surface-exposed loops has become accessible to externally added protease.This result indicates that the exchanged fragment between residues 423 and 479 affects the barrel structure and probably belongs to the β-barrel domain, implying that this domain contains 16 β-strands.
To test whether individual 16-or 12-stranded NmBamA βbarrels could be stably inserted in the OM, we produced such proteins, designated 420 NmBamA (mature protein starting with amino acid 420 of NmBamA) and 481 NmBamA (mature protein starting with amino acid 481 of NmBamA), in HB-1.Urea extraction experiments indicated that 420 NmBamA was completely and 481 NmBamA was only partially inserted in the OM (Figure 8B).Trypsin treatment of cell envelopes showed that 420 NmBamA was completely protected from the protease, whereas 481 NmBamA was for the most part degraded (Figure 8C).The 420 NmBamA protein migrated at exactly the same position as the trypsin-protected fragment of Ec 423 Nm, suggesting that they possess a similar membrane domain, i.e. a stably integrated 16-stranded β-barrel.

Discussion
In this work we studied the species-specific functioning of BamA, the central component of the OMP insertion machinery, which is essential and conserved in Gram-negative bacteria.Previously, it was reported that the BamA protein of the cyanobacterium Anabaena sp.PCC 7120 did not function in E. coli, despite its insertion in the E. coli OM [31].This lack of function may be due to the large phylogenetic distance between these bacterial species which is reflected in the distinct pore diameters, differing number of POTRA domains (3 versus 5) and differences in associated complex components for BamA proteins from cyanobacterial and proteobacterial origin [32][33][34][35].We now show that the BamA species specificity extends to much more closely related species, since even BamA homologs of proteobacteria or even within one class of proteobacteria were generally not exchangeable.
Several factors might potentially contribute to this lack of cross-complementation: insufficient assembly of the heterologous BamA into the OM of the host, an inability to associate with the accessory lipoproteins into a fully functional OMP-insertion machinery, or failure of the heterologous BamA to efficiently recognize various OMP substrates.E. coli BamA was not assembled into the OM of N. meningitidis and therefore could obviously not compensate for the absence of neisserial BamA.In contrast, N. meningitidis BamA appeared to be inserted into the E. coli OM, at least to a substantial extent.Previously, we showed that neisserial OMPs are only poorly inserted into the E. coli OM.This defect was related to the nature of the penultimate amino acid of the OMP: in Neisseria this residue is almost invariably a positively charged one, whereas in E. coli this is only very rarely the case [23].Replacing the lysine present in this position in the neisserial porin PorA by glutamine greatly enhanced PorA assembly in E. coli [23].NmBamA is one of the few neisserial OMPs possessing a threonine instead of a positively charged residue at the penultimate position; therefore, it might be sufficiently recognized by the Bam complex to be inserted into the E. coli OM.Neisserial BamA was even able to associate in the E. coli OM with the essential Bam-complex component BamD but not with BamB.The latter is consistent with the absence of a BamB homolog in N. meningitidis [14].We did not test association of BamC and BamE, but, since these proteins have been shown to interact indirectly with BamA via BamD [13,36], they are likely to form part of the hybrid complex.Thus, despite its proper assembly into the OM and its association with at least the other essential Bam complex component, NmBamA could not functionally replace EcBamA.Therefore, the most likely explanation for its lack of function in E. coli is poor recognition of the host's OMP substrates.
Paradoxically, whereas the lack of substrate recognition is apparently an issue when BamA proteins are exchanged between different proteobacteria, we recently showed that E. coli OMPs could be assembled into the OM of yeast mitochondria [25], and vice versa, that a mitochondrial OMP could be inserted into the E. coli OM [26].However, these assemblies were efficient only when the OMPs were expressed at a low level with extensive amounts of unassembled OMPs accumulating at higher expression levels.It seems likely that also NmBamA can handle E. coli OMP substrates to some extent.However, due to the fact that for cellular survival it needs to deal with all OMPs under conditions of suboptimal recognition, it is not unexpected that NmBamA fails to sufficiently handle this load of substrates.Cells then likely die from lethal accumulation of periplasmic OMP aggregates.
Chimeric proteins comprising BamA domains derived from N. meningitidis and E. coli were not functional in either host.All evidence indicated that the two chimeras containing the Cterminal part of N. meningitidis BamA were inserted into the neisserial OM.However, the conformation of the β-barrel domain of the Ec 479 Nm protein was different as indicated by its accessibility to trypsin in intact cells.The apparent cleavage in one of its extracellular loops also explains the considerably smaller trypsin-resistant fragment detected after trypsinization of the Ec 479 Nm protein in cell envelopes (Figure 6B).In contrast, similar to wild-type NmBamA, the Ec 423 Nm chimera was resistant to trypsin digestion in intact cells.The most straightforward explanation for these results is that the protein segment between residues 423 and 479 does not constitute a hinge region between the POTRA domains and the β-barrel according to the original 12-stranded β-barrel model, but is actually part of the β-barrel.Consequently, in the Ec 479 Nm protein, four β-strands of the NmBamA β-barrel are substituted by the corresponding ones of EcBamA and this substitution results in a conformational change rendering a surfaceexposed loop near the C-terminal end sensitive to trypsin.Overall, these data support a 16-stranded rather than the originally proposed 12-stranded membrane β-barrel.Interestingly, during submission of this manuscript the structure of N. gonorrhoeae BamA, refolded from solubilized inclusion bodies, was elucidated [37] and found to consist of a large periplasmic domain attached to a 16-stranded β-barrel domain.Thus, our observations from in vivo assembled proteins are consistent with the reported BamA structure.
Despite apparently correct membrane insertion of the Ec 423 Nm and Ec 479 Nm chimeras, the proteins could not functionally replace NmBamA.Trypsin digestion of the chimeras in cell envelopes did not yield the intermediate fragments seen for native BamA.These fragments are indicative for association with other Bam complex members [14].Therefore, the EcBamA-derived POTRA domains in both chimeras apparently fail to associate with the accessory Bam complex components in N. meningitidis, which would explain the lack of function of both chimeras in N. meningitidis.In

Figure 1 .
Figure 1.Expression and assembly of E. coli BamA in N. meningitidis.Cell envelopes were analyzed by SDS-PAGE followed by immunoblotting using anti-EcBamA antiserum or by staining with Coomassie Brilliant Blue.BamA is indicated with arrowheads.A: Immunoblots of cell envelopes of uninduced (-) or induced (+) HB-1 cells containing pFP10-EcbamA carrying bamA under an IPTG-inducible promoter.B: Cell envelopes of HB-1 (Nm) expressing EcBamA or of E. coli strain DH5α (Ec) were extracted with urea and the input (i), urea-insoluble (p) and -soluble (s) fractions were analyzed by SDS-PAGE followed by staining with Coomassie Brilliant Blue (lower panels) or immunoblotting with anti-EcBamA antiserum (upper panels).Neisserial porins are indicated with asterisks.C: Cell envelopes of E. coli strain DH5α (Ec) or HB-1 (Nm) expressing EcBamA were treated (+) or not (-) with trypsin for 1 h and analyzed by SDS-PAGE followed by immunoblotting with anti-EcBamA antiserum.Trypsin-protected EcBamA fragments in E. coli are indicated with an asterisk.doi: 10.1371/journal.pone.0085799.g001

Figure 2 .
Figure 2. Functionality of BamA variants in E. coli.Growth of E. coli strain UTP BAD ::bamA carrying no plasmid (panel A) or plasmids encoding the proteins indicated in the upper right hand side of the graphs was assessed by measuring the OD 600 .Strains were grown in the presence of glucose (open circles), arabinose (closed diamonds) or IPTG (closed triangles) in LB at 37°C.Cultures were diluted into fresh medium at the time points indicated by arrows.doi: 10.1371/journal.pone.0085799.g002

Figure 3 .
Figure 3. Expression and assembly of N. meningitidis BamA in E. coli.A: Expression of NmbamA in E. coli.E. coli strain UTP BAD ::bamA containing pFP10-NmbamA was grown in LB containing 0.02% arabinose in the absence or presence of IPTG.Cell envelopes were isolated and analyzed by immunoblotting with Mab α-POTRA1 Nm .B: Assembly of NmBamA in the E. coli OM.Cell envelopes from E. coli strain UTP BAD ::bamA containing pFP10-NmbamAgrown in LB containing 0.02% arabinose and IPTG were extracted with urea (Ec).As a control, cell envelopes of strain HB-1 were extracted with urea (Nm).Input (i), urea-insoluble (p) and -soluble (s) fractions were analyzed by SDS-PAGE and immunoblot analysis with Mab α-POTRA1 Nm .C: Co-purification of Bamcomplex components with His-tagged NmBamA in E. coli.Cell envelopes of E. coli DH5α cells containing pRV-His-NmbamA either induced (lanes 2) or not (lanes 1) with IPTG were extracted with Elugent and subjected to Ni 2+ -NTA purification.Shown are elution fractions analyzed by denaturing SDS-PAGE and silver staining (left panel) or immunoblotting (right panels) using antisera against the indicated proteins.As a positive control for BamB detection, cell envelopes derived from strain DH5α were also analyzed on blot (lane c).The arrow indicates the position of BamB.doi: 10.1371/journal.pone.0085799.g003

Figure 4 .Figure 5 .
Figure 4. Expression of heterologous bamA variants in N. meningitidis and E. coli.A, B: N. meningitidis strain HB-1 and its derivatives were grown with IPTG to induce expression of B. pertussis (Bper), B. mallei (Bmal), N. gonorrhoeae (Ngo) and as control, E. coli (Eco) bamA genes from plasmids.A: Cell lysates were analyzed by SDS-PAGE and immunoblotting with the antisera indicated on the right.Note that also chromosome-encoded NmBamA is detected with the α-N-NmBamA antiserum in the upper panel.B: RNA was isolated and treated or not with reverse transcriptase (RT).Specific cDNA was detected by conventional PCR followed by agarose gel electrophoresis.The -RT samples serve as controls for the absence of plasmid DNA in the RNA preparations.C: RNA was isolated from UTP BAD ::bamA containing pFP10-BmbamA (Bmal) or pFP10-BpbamA (Bper) grown in the presence of arabinose and IPTG and processed as explained panel B. doi: 10.1371/journal.pone.0085799.g004

Figure 6 .
Figure 6.Expression and assembly of BamA chimeras in N. meningitidis.A: Derivatives of strain HB-1 containing plasmids encoding the chimeric proteins indicated above the panels were grown with or without IPTG.Cell envelopes were isolated and analyzed by SDS-PAGE and immunoblotting with the antisera shown below the blots.The chimeric proteins are indicated with asterisks.Note that also chromosome-encoded NmBamA is detected with the α-N-NmBamA antiserum (indicated with the arrowhead).B: Trypsin sensitivity of BamA variants in N. meningitidis cell envelopes.Cell envelopes of strain HB-1 expressing the indicated BamA variant were treated overnight with or without trypsin and analyzed on immunoblots.Chromosomally encoded NmBamA in the parent strain HB-1 (wt) yielded three tryptic fragments indicated as I, II and III.Additional digestion products of Ec 479 Nm and Ec 423 Nm are indicated with asterisks.Arrowheads indicate the positions of undigested NmBamA and the Ec 423 Nm and Ec 479 Nm chimeras, which all have indistinguishable electrophoretic mobilities.In the righ-hand panel, asterisks indicate the positions of undigested Nm 423 Ec and Nm 480 Ec.The blot required a long exposure time to visualize Nm 480 Ec which additionaly revealed a cross-reactive band just below the signal of the chimeras.C: Extractability of BamA variants from cell envelopes with urea.Cell envelopes of N. meningitidis strain HB-1 producing the chimeras indicated at the right were extracted with urea and the input (i), urea-insoluble (p) and -soluble (s) fractions were analyzed by SDS-PAGE and immunoblotting with the antibodies indicated on the left.Arrowheads indicate chromosomally encoded endogenous NmBamA.In the lower two panels, the positions of endogenous BamA and the chimeric proteins are indistinguishable.doi: 10.1371/journal.pone.0085799.g006

Figure 7 .
Figure 7. Assembly of BamA chimeras in E. coli.Derivatives of E. coli strain UTP BAD ::bamA containing plasmids encoding BamA variants were grown in LB containing arabinose and IPTG.Cell envelopes were isolated and extracted with urea.Input (i), urea-insoluble (p) and -soluble (s) fractions were analyzed by SDS-PAGE and immunoblotting with the antibodies indicated on the left.The BamA variants encoded by the plasmids are indicated at the right.doi: 10.1371/journal.pone.0085799.g007

Figure 8 .
Figure 8. Analysis of the β-barrel domain of NmBamA.A: Intact HB-1 cells and HB-1 cells producing Ec 423 Nm or Ec 479 Nm were treated with the indicated concentrations of trypsin for 15 min and processed for immunoblotting with α-POTRA1 Nm (left panels) or α-EcBamA (right panels) antibodies.B. Cell envelopes of HB-1 producing the proteins indicated at the right were treated with urea and processed for immunoblotting with α-C-NmBamA antiserum.C: Cell envelopes of HB-1 producing the proteins indicated on top were treated overnight with or without trypsin and analyzed on immunoblots with α-C-NmBamA antiserum.The # signs indicate the positions of the truncated 420 Nm BamA and 481 NmBamA variants, wheras the * signs indicate tryptic fragments derived from the chimeric Ec 423 Nm or Ec 479 Nm proteins (similar to those shown in Fig. 6B).doi: 10.1371/journal.pone.0085799.g008