Mapping the Laminin Receptor Binding Domains of Neisseria meningitidis PorA and Haemophilus influenzae OmpP2

Neisseria meningitidis, Haemophilus influenzae and Streptococcus pneumoniae are major bacterial agents of meningitis. They each bind the 37/67-kDa laminin receptor (LamR) via the surface protein adhesins: meningococcal PilQ and PorA, H. influenzae OmpP2 and pneumococcal CbpA. We have previously reported that a surface-exposed loop of the R2 domain of CbpA mediates LamR-binding. Here we have identified the LamR-binding regions of PorA and OmpP2. Using truncated recombinant proteins we show that binding is dependent on amino acids 171–240 and 91–99 of PorA and OmpP2, respectively, which are predicted to localize to the fourth and second surface-exposed loops, respectively, of these proteins. Synthetic peptides corresponding to the loops bound LamR and could block LamR-binding to bacterial ligands in a dose dependant manner. Meningococci expressing PorA lacking the apex of loop 4 and H. influenzae expressing OmpP2 lacking the apex of loop 2 showed significantly reduced LamR binding. Since both loops are hyper-variable, our data may suggest a molecular basis for the range of LamR-binding capabilities previously reported among different meningococcal and H. influenzae strains.


Introduction
Invasive disease caused by Neisseria meningitidis (meningococcus), Haemophilus influenzae and Streptococcus pneumoniae (pneumococcus) remains an important cause of morbidity and mortality worldwide, despite the use of antibiotics and the introduction of effective conjugate polysaccharide vaccines that target some diseaseassociated serogroups [1]. However, there is no currently licensed vaccine effective against serogroup B meningococcus, the dominant serogroup in much of the developed world [2]. Similarly, there are no vaccines against non-type B H. influenzae and the vaccines available against pneumococcus only provide protection against a limited subset of the plethora of recognized serogroups [1].
The human 37/67-kDa laminin receptor (LamR) is a highly conserved, multi-functional protein [3]. The relationship between the 37-kDa and 67-kDa forms is not completely understood, but the former is thought to mature into one or more homo or heterodimeric 67-kDa forms [4,5,6]. LamR was initially identified as a cell surface receptor for the extracellular matrix molecule laminin [7,8,9]. LamR is important for cell adhesion to the basement membrane and is also implicated in tumour cell metastasis [10]. LamR has additional roles in intracellular signaling [11], ribosomal activity [12] and cell viability [13]. LamR can also migrate to the nucleus where it can bind the histones H2A, H2B and H4, although the significance of this is not known [14].
Recently we showed that binding to LamR is a common mechanism employed by S. pneumoniae, N. meningitidis and H. influenzae to adhere to human brain microvascular endothelial cells [22]. The bacterial ligands responsible for LamR binding were identified as pneumococcal CbpA, meningococcal PilQ and PorA and H. influenzae OmpP2 [22]. All are abundant, multi-functional proteins with surface-exposed loop structures. In particular, meningococcal PorA and H. influenzae OmpP2 share many characteristics, despite showing limited sequence similarity. Both are homo-trimeric outer membrane porins with amphipathic bbarrel structures forming sixteen membrane-spanning b-strands separating more variable sequences forming eight surface exposed loops [23,24]. In the case of OmpP2, most sequence variability occurs in the second, fourth, fifth and eighth loops (known as variable regions 1, 2, 3 and 4, respectively) [25,26], whereas in PorA most variation occurs in the first, fourth and fifth loops (variable regions 1, 2 and 3, respectively) [27,28]. Indeed sequence variability of PorA is the basis of the meningococcal subtyping strategy [29,30,31].
As immunodominant antigens and targets for serum bactericidal antibodies, PorA and OmpP2 have been studied as potential vaccine targets [32,33]. However, bactericidal antibodies elicited by such vaccines predominately target the variable extra-cellular loops [23,27,34]. Therefore, generating an immune response against one antigen does not generally confer protection against strains with heterologous antigens.
Previously, we showed that the highly conserved surfaceexposed loop (residues 391 EPRNEEK 397 ) linking the second and third anti-parallel a-helices of the R2 domain of pneumococcal CbpA mediated LamR-binding [22]. Here, using recombinant derivatives of PorA and OmpP2, synthetic peptides corresponding to their extracellular loops, and bacterial strains expressing LamR ligands in which specific extracellular loops were deleted, we identify the fourth and second extra-cellular loops of PorA and OmpP2, respectively, as the LamR-binding domains of these proteins.
Next, derivatives of PorA 20-392 were expressed in which the residues spanning loop 4 were removed (PorA D169-240 ). This derivative showed significantly less LamR-binding activity than PorA   (Fig. 1D). In contrast, control PorA derivatives lacking parts of loops 1 (PorA D58-94 ) or 3 (PorA D125-149 ) exhibited similar LamR binding to PorA 20-392 , providing additional confirmation that these regions are not required for optimal LamR binding ( Fig. 1D). Taken together these results indicated that the LamRbinding domain of MC58 PorA lies within amino acids 171-240 and was likely localized to the fourth extra-cellular loop which comprises amino acids 185-218.
We hypothesized that the LamR-binding activity was localized to one of the extra-cellular loops of OmpP2 between amino acids 24-224; four derivatives of OmpP2 24-359 were expressed and purified in which the residues spanning the apical regions of loops 1, 2, 3 or 4 were removed. The derivative lacking loop 2 (OmpP2 D91-99 ) showed significantly less LamR binding activity compared to OmpP2   (Fig. 2C), whilst the other derivatives lacking loops 1, 3 or 4 showed no significant reductions in binding. These results indicated that the LamR-binding domain of OmpP2 lies within amino acids 91-99, corresponding to the apical region of the second extra-cellular loop, which comprises amino acids 88-102.
Synthetic peptides corresponding to PorA loop 4 and OmpP2 loop 2 bind LamR and block ligand-LamR binding Further characterization of the PorA and OmpP2 LamRbinding domains was undertaken using synthetic peptides. Several peptides were synthesized (Table 1): one corresponding to the 9amino acid apical region of OmpP2 loop 2 (residues 91-99); another corresponding to the 34-amino acids of PorA loop 4 (residues 185-218), which we hypothesized was responsible for PorA-LamR binding; and a third corresponding to PorA loop 1 (residues 31-70) for use as a non-LamR binding control peptide. In addition, scrambled peptides based on OmpP2 loop 2 and PorA loop 4 were also synthesized. Each peptide included additional terminal cysteine residues; under oxidizing conditions cysteine sulfhydryl groups spontaneously form disulfide bonds, thus allowing the peptides to circularize and more closely mimic the native loop structures. PorA loop 1 and the two scrambled peptides (PorA loop 4scr and OmpP2 loop 2scr) bound less than either OmpP2 loop 2 or PorA loop 4 to LamR (Fig. 3A). Loop 4 residues 192-208 are required for optimal LamRbinding Experiments were undertaken to further define the region within the 34-amino acid PorA loop 4 required for optimal LamR binding. Five overlapping peptides (L4_1 to L4_5) were synthesized spanning loop 4 ( Table 1 & Fig. 4A). The peptides were immobilized in microtitre wells and DIG-labeled LamR added. After washing, bound LamR was detected and quantified (Fig. 4B). As expected, the control non-LamR binding control peptide (loop 1) showed significantly less binding to LamR compared to whole loop 4. Two of the shorter loop 4 peptides (L4_4 and L4_5) showed significantly reduced LamR-binding compared to whole loop 4 (p,0.05), whilst L4_1 showed reduced binding, albeit it not reaching statistical significance (p = 0.069). The highest LamRbinding was exhibited by L4_2 and also L4_3, which correspond to the most apical part of loop 4 and together span PorA 192-208 suggesting that the key residue(s) mediating LamR-binding are localized in this region.

Deletion of N. meningitidis PorA loop 4 or H. influenzae OmpP2 loop 2 reduces LamR-binding
To determine whether the LamR-binding activity of PorA loop 4 observed using synthetic peptides and recombinant subfragments reflect the ability of the protein in situ in the bacterial cell to mediate binding to LamR, a derivative of the N. meningitidis strain MC58 was constructed in which the central 14 amino acids of PorA loop 4 were deleted (MC58PorA D197-210 ). We have previously shown that meningococci have two LamR-binding ligands: PorA and PilQ, and that in the presence of PilQ, deletion of PorA has relatively little effect on observable LamR-binding [22]. Therefore, we introduced a pilQ mutation into the MC58PorA D197-210 genetic background to create MC58PorA D197-210 DpilQ to enable us to specifically address the effect of loop 4 truncation on PorA-mediated LamR-binding. Truncating loop 4 did not interfere with expression of PorA or the ability of the protein to correctly localize to the meningococcal outer membrane. Immunoblots of cell fractions confirmed the presence of PorA D197-210 at similar levels to wild-type PorA in meningococcal outer membrane-enriched sub-cellular fractions (Fig. 5A). In addition, a mouse monoclonal antibody directed against loop 1 bound to the surface of strains expressing wild-type PorA or PorA D197-210 at similar levels as observed by confocal microscopy, but not in strains where porA had been inactivated (Fig. 6). LamR-binding exhibited by MC58PorA D197-210 DpilQ was significantly reduced and similar to the control strain lacking PorA and PilQ (Fig. 7A), thus confirming that the apical region of loop 4 determines PorA-mediated LamR binding by N. meningitidis MC58.
Similarly, we constructed a H. influenzae Rd KW20 derivative in which the amino acids 91-99 were deleted (RdOmpP2 D91-99 ). Again, prior to determining the LamR-binding potential of this strain, we first confirmed that truncating loop 2 did not interfere with expression of OmpP2 or the ability of the protein to correctly localize to the H. influenzae outer membrane (Fig. 5B). LamRbinding by RdOmpP2 D91-99 was significantly reduced compared to the wild-type (p,0.01), albeit the reduction in LamR-binding was not as great as that observed for the control strain lacking OmpP2 (Fig. 7B), perhaps indicating that additional residues of OmpP2 may play a role in optimizing LamR-binding. Control strains with an antibiotic cassette inserted at the porA or ompP2 locus, but no modification to either the porA or ompP2 coding sequence, showed no significant reduction in LamR-binding, thus confirming the direct effect of removing or truncating porA or ompP2 on binding (data not shown).
In summary, we have identified the apical regions of the fourth and second extra-cellular loops of meningococcal MC58 PorA and H. influenzae Rd KW20 OmpP2, respectively, as the LamRbinding regions of these bacterial ligands.

Discussion
Several neurotropic pathogens bind to the 37/67-kDa laminin receptor on host microvascular endothelial cells. In particular, this adherence strategy is used by the bacterial pathogens S. pneumoniae, N. meningitidis and H. influenzae [22]. These species target a common carboxy-terminal recognition site on LamR (amino acids 263-282), since antibodies recognizing this sequence or a peptide corresponding to LamR residues 263-282 could inhibit bacterial binding to microvascular endothelial cells [22]. The bacterial ligands required for LamR-binding were subsequently identified as pneumococcal CbpA, meningococcal PilQ and PorA and H. influenzae OmpP2 [22]. Understanding the structural basis for the ability of these ligands to bind LamR could facilitate the design of therapeutic interventions which could prevent or disrupt the interaction and thus engender protection against bacterial meningitis.
To this end, we previously investigated the LamR-binding region of CbpA: a secreted protein that acts as a bridging adhesin to human cells [35]. The CbpA LamR-binding region mapped to a highly conserved surface-exposed loop (residues 391 EPR-NEEK 397 ) linking the second and third anti-parallel a-helices of the R2 domain. Pneumococcal strains which failed to bind LamR were shown to have this region missing from CbpA [22]. Two residues within this region were found to be particularly important for LamR binding; pneumococci expressing a CbpA P392G -R393G derivative showed significantly reduced binding to recombinant LamR and endothelial cells. CbpA P392G -R393G -coated microspheres showed significantly reduced binding to mouse cerebral endothelium compared to those bearing wild-type CbpA [22]. Importantly, in a mouse model of pneumococcal meningitis, strains expressing CbpA P392G -R393G rarely caused meningitis compared to strains expressing wild-type CbpA, thus underlining the in vivo significance of these key residues in the disease process in the mouse model [22].
In this study, we undertook to define the regions of meningococcal PorA and H. influenzae OmpP2 required for LamR-binding. Both are abundant outer membrane trimeric porins that contain multiple surface-exposed loop structures, but lack sequences resembling the CbpA LamR-binding domain. Using sub-fragments of recombinant MC58 PorA and Rd KW20 OmpP2, our results suggest that their LamR-binding domains are localized to amino acids 171-240 and amino acids 91-99 respectively, since derivatives containing these regions bound LamR; in contrast derivatives lacking these regions showed significantly reduced LamR-binding. For both PorA and OmpP2, predicted surfaceexposed loops mapped to these LamR-binding regions. The importance of these loops was confirmed using synthetic peptides based on their sequences. Peptides corresponding to either PorA loop 4 or OmpP2 loop 2 bound LamR, whilst scrambled peptides based on these loops or a peptide based on PorA loop 1 bound less well. Additionally, PorA loop 4 could significantly inhibit binding of LamR to full-length PorA in contrast to PorA loop 1.
To further refine the region within the 34-amino acid loop 4 required for optimal LamR-interaction, the binding abilities of shorter overlapping peptides spanning the loop were determined. This approach was not undertaken for OmpP2 loop 2, since our data already indicated that optimal binding was dependant on a short 9-amino acid region (residues 91-99) corresponding to the apex of the loop. Likewise for PorA loop 4, the highest LamRbinding was exhibited by short peptides mapping to the loop apex, suggesting that the key residue(s) mediating binding are localized to this region.
LamR-binding exhibited by synthetic peptides or ligand subfragments purified under non-native conditions might not accurately reflect the ability of native ligands embedded in the Gram negative outer membrane to interact with LamR. We therefore confirmed our findings using H. influenzae and meningococcal LamR-ligands expressed in their in vivo context. H. influenzae Rd KW20 expressing OmpP2 containing a 9-amino acid deletion from the apex of loop 2 showed significantly reduced LamR-binding compared to wild-type. However, LamR-binding was not reduced to the level determined for the H. influenzae DompP2 control strain, indicating that residues outside of the deleted region may also have a minor effect to modulate binding. Recent investigations into the role of OmpP2 and its loops in pathogenesis have focused on the ability of the protein and some of its loops (principally loop 7) to activate host signaling cascades (including the mitogen-activated protein kinase cascade) and initiate the innate immune response (inducing the release of TNF-alpha and IL-6) in a variety of human cell types [36,37,38,39]. Here we show that, in addition to the demonstrated role of OmpP2 loop 7 in pathogenesis, loop 2 is also biologically relevant as it largely determines the LamR binding activity of OmpP2.
Our analysis of PorA-mediated LamR-binding in the meningococcus was complicated by the presence of PilQ, the second LamR ligand expressed by this organism. Consequently, our LamR-binding experiments were undertaken in pilQ knockout  derivatives of MC58 to exclude the effects of this protein. Importantly, N. meningitidis expressing PorA lacking 14-amino acids from the apex of loop 4 exhibited a significant reduction in LamR-binding compared to strains expressing wild-type PorA. In fact, removing these amino acids from PorA in a pilQ knockout background, reduced LamR-binding to levels indistinguishable from that of a pilQ porA double knockout mutant. Importantly, this was not due to the inability of the modified PorA to insert into the meningococcal outer membrane. Based on our data, we conclude that the fourteen amino acids of loop 4 represent the sole LamRbinding domain of PorA in this strain. Further deletions or sitedirected mutagenesis of residues in this region of PorA may lead to further refinement of this binding domain. Loop 2 corresponds to variable region 1 of OmpP2 [24]. Thus, antigenic variation at this loop is likely to explain the varying LamR-binding capabilities of H. influenzae isolates reported previously [22]. Similarly, the fourteen amino acid LamR-binding region of PorA corresponds to the VR2 of this antigen. Our identification of VR2 as the LamR-binding of PorA in this study is consistent with previous observations from an experiment in which porA of a meningococcal isolate with a low LamR-binding activity (Z4682) was replaced with porA from MC58 [22]. Strain Z4682 bound LamR at 50% of the level of MC58; following the porA replacement, LamR-binding was increased to 94% of MC58 [22]. The PorA variants (VR1, VR2) of MC58 and Z4682, respectively, are: 7, 16-2 and 7.1, 1 (http://pubmlst.org/neisseria/PorA/). Differences in VR1 sequence between variants 7 and 7.1 are minor, involving the presence/absence of a duplication of two residues. In contrast, the differences in amino acid sequence between VR2 variants 16-2 and 1 are considerable (16-2: YYTKNTNNNLTLVP; 1: YVAVENGVAKKVA). Since VR1 corresponds to loop 1, and VR2 to loop 4, this provided circumstantial evidence linking loop 4, but not loop 1 to differences in LamR-binding ability. Here, using a comprehensive approach involving PorA sub-fragments, synthetic peptides and meningococcal strains expressing PorA in which specific extracellular loops were deleted, we confirm that the fourth extra-cellular loop is the LamR-binding region of this protein.
Since over 630 PorA VR2 variants (classified into 20 families) have been identified across meningococcal isolates to date, we suggest that the LamR-binding ability conferred by PorA is variable between strains. In contrast, since PilQ is highly conserved ($98% identity at amino acid level between genome sequenced meningococcal strains), we hypothesize that PilQ confers a relatively constant LamR-binding capability to strains of different lineages. We are currently determining the LamRbinding domain(s) of PilQ to allow us to investigate this further. Why the meningococcus has evolved to express two LamRbinding ligands, whilst H. influenzae and pneumococcus both express only one is unclear. However, meningococci have previously been reported to express multiple adhesins with overlapping ligand-binding capabilities [40]. Functional redundancy is required to maintain colonization because of the propensity of the organism to phase and/or antigenically vary its surface structures. PorA is one such variable surface structure; in addition to being antigenically variable, PorA expression levels vary between isolates and within a population of the same isolate. This is a result of slipped-strand mispairing at a homopolymeric tract in the porA promoter [41]. In addition, PorA expression may be absent altogether due to: slipped-strand mispairing at a homopolymeric tract within the porA gene [41]; deletion of porA [42]; or via the insertion of an IS element [43]. Whether PilQ  expression is varied in similar ways and if so, what the genetic mechanisms governing this variation are, are not currently known. However, we suggest that the meningococcus utilizes two LamRligands in order to maintain a degree of LamR-binding capability, irrespective of changes in PorA (or possibly PilQ) expression, during the infection process. Since meningococci are normally commensal organisms inhabiting the nasopharynx, and infection of the meninges represents a relatively rare event that is unlikely to contribute to the dissemination of this accidental pathogen, selective pressure leading to such functional redundancy is likely to be applied at mucosal surfaces and not at the meninges.
In summary, we have identified the LamR-binding regions within meningococcal PorA and H. influenzae OmpP2. These regions correspond to the apical regions of the fourth and second extra-cellular loops, respectively. Both loops are hyper-variable, suggestive of the molecular basis for the diverse range of LamRbinding capabilities previously reported for meningococcal and H. influenzae strains of different lineages. Increased knowledge of the structural motifs of bacterial ligands that interact with LamR may facilitate the design of therapeutic interventions which could disrupt or modulate the interaction of neuroinvasive bacteria with LamR and engender protection against bacterial meningitis. However, given that both OmpP2 loop 2 and PorA loop 4 are hyper-variable, this may hinder the design of strategies to block LamR-binding, which are effective against a wide range of pathogenic strains.

Bacterial strains and growth conditions
Escherichia coli JM109 was used as the host strain for mutagenic constructs and for expression of 66 histidine-tagged recombinant protein fragments (Table S1). This strain was grown at 37uC in  Lysogeny Broth (LB) or on LB agar supplemented, where appropriate, with ampicillin, kanamycin or streptomycin and spectinomycin (all at 100 mg ml 21 ). N. meningitidis strains were grown at 37uC, in an atmosphere of air plus 5% CO 2 , on Columbia agar with chocolated horse blood (Oxoid) or in Brain Heart Infusion (BHI) broth (Oxoid) supplemented, where appropriate, with streptomycin and spectinomycin (100 mg ml 21 ) and/ or kanamycin (50 mg ml 21 ). H. influenzae strains were grown at 37uC, in an atmosphere of air plus 5% CO 2 , on Columbia agar with chocolated horse blood (Oxoid) or in BHI broth supplemented with haemin (10 mg ml 21 ) and NAD (10 mg ml 21 ), and where appropriate, with kanamycin (50 mg ml 21 ).

DNA manipulation
Chromosomal DNA was purified using the DNeasy Tissue kit (Qiagen). Plasmid DNA was prepared by using the QIAprep Spin kit (Qiagen). Restriction enzymes were purchased from Roche and used according to the manufacturer's instructions. DNA sequencing was carried out on an ABI 377 automated DNA sequencer at the School of Biomedical Sciences (University of Nottingham).

Protein expression and purification
Recombinant histidine-tagged LamR was purified as described previously [22]. For the expression of PorA and OmpP2 derivatives, plasmids were transformed into E. coli JM109, cultures grown to exponential phase (OD 600 .0.5) and induced with isopropyl b-D-1-thiogalactopyranoside (IPTG; 1 mM) for 3 h. Recombinant proteins were then affinity-purified using HisPur TM Cobalt Resin (Pierce) under denaturing conditions using a previously described method [44]. Final column elutes were buffer exchanged into 16 PBS using PD-10 desalting columns (GE Healthcare) according to the manufacturer's instructions.

DIG-labeling of proteins and bacteria
DIG-labeling was carried using the Roche DIG-NHS Protein labeling kit, used as per manufacturer's instructions. Briefly, for protein the pH was adjusted to pH 9.0 by addition of 2 M sodium carbonate, and DIG-NHS reagent added in a 1:10 molar ratio. Reactions were incubated at room temperature for 1 h, and then unbound DIG-NHS was removed using PD-10 desalting columns (GE Healthcare). For bacteria, cells were harvested from an overnight plate, washed 3 times in PBS with 0.05% Tween 20 (PBST) and resuspended in sodium carbonate buffer (150 mM; 142 mM NaHCO 3 , 8 mM Na 2 SO 3 , pH 9.0). Bacteria (optical density at 600 nm = 1.0) were then labeled with 10 mg ml 21 DIG-NHS at room temperature for 30 min. Bacteria were then washed 3 times with PBST and resuspended in PBS containing 1% bovine serum albumin (BSA; Sigma).

Enzyme-linked immunosorbent assays
100 ml aliquots of 5 mg ml 21 protein in sodium carbonate buffer were added to a CovaLink NH microplate (Nalge Nunc International, USA), and incubated at room temperature for 1 h or overnight at 4uC. The plate was washed three times with PBST and blocked with 1% BSA/PBS for 1 h. After removal of the blocking solution, 100 ml of 5 mg ml 21 DIG-labeled protein or DIG-labeled bacteria (OD 600 at 0.04) was added and incubated at room temperature for 90 min or at 4uC overnight. Following vigorous washing in PBST (5 washes with 5 min soaking times), 100 ml of anti-DIG HRP conjugate (Roche), diluted 1:5000 in 1% BSA/PBS, was added to wells and incubated for 1 h. Plates were again vigorously washed and color developed by adding 100 ml ABTS substrate (Roche). Plates were read with an ELISA reader (Biotek EL800) at an absorbance of 405 nm. Alternatively, for unlabelled LamR, 100 ml of protein (5 mg ml 21 ) was added to coated wells. After washing, bound LamR was detected using rabbit polyclonal anti-LamR (diluted 1:1000) in 1% BSA/PBS followed by goat anti-rabbit IgG-HRP conjugate (Sigma; diluted 1:10,000). Inhibition assays were performed as described above, except that 0.5 mg aliquots of DIG-labeled LamR were preincubated with 0.5-100 mg of peptide for 4 h at room temperature prior to being added to coated wells. OD values shown are minus OD values obtained from control wells coated with 100 ml aliquots of 1% BSA. Statistical significance was determined using twotailed Student's t test.

Synthetic peptides
The synthetic PorA and OmpP2 peptides utilized in this study are detailed in Table 1. All were synthesized (high purity grade; peptide purity .95%) by GenScript, USA.

PorA loop 4 truncation in N. meningitidis
A 42-bp deletion in the coding sequence of MC58 porA was introduced into a previously constructed plasmid, pPorA-4d (Table S1), by inverse PCR using primers L4-F1 and L4-R1 (Table S2). Amplification resulted in a product containing the desired deletion and incorporating unique BglII and BamHI sites, thus allowing re-ligation of the fragment to form pPorADL4 (Table  S1). Inverse PCR using primers L4-F2 and L4-R2 was then used to introduce a BglII site downstream of porA, into which a BglIIdigested V cassette (encoding resistance to spectinomycin and streptomycin) [45] was inserted to form pPorADL4V (Table S1). This plasmid was subsequently used to mutate N. meningitidis MC58 by natural transformation and allelic exchange as previously described [46]. The porA loop 4 deletion in the resulting mutant (MC58PorA D197-210 ) was confirmed by PCR analysis and DNA sequencing. Growth curve assays carried out using liquid cultures showed no significant differences between MC58PorA D197-210 and the wild-type strain (data not shown). The pilQ knockout mutation was subsequently introduced into this strain to generate MC58PorA D197-210 DpilQ using previously described methods [22].

OmpP2 loop 2 truncation in H. influenzae
A 2.7-kb region containing the ompP2 gene and flanking DNA was amplified from H. influenzae Rd KW20 using primers F1 and mP2_R (Table S2) and TA-cloned into pGEM-T Easy to yield pMSA8 (Table S1). This template was subject to inverse PCR using OmgF and OmgR primers, thus incorporating BamHI restriction sites into the amplicon and a copy of a DNA uptake sequence (59-AAGTGCGGTCA-39) which is required for efficient DNA uptake by H. influenzae [47]. The BamHI site was used to introduce a BamHI-digested kanamycin resistance cassette downstream of, and in the same orientation as ompP2, to generate pMSA15. Inverse PCR utilizing primers P2DL2I_F and P2DL2I_R was subsequently used to remove the nucleotides encoding OmpP2 amino acids 91-99. Use of P2DL2I_F and P2DL2I_R incorporated BglII-sites into the amplicon, thus allowing re-ligation to form pMSA16. Nucleotide sequencing revealed a single base pair deletion present in the 59 end of the ompP2 coding sequence in pMSA8 and in subsequent constructs which would lead to premature translation of ompP2 in H. influenzae. To repair this, a 539-bp mega primer spanning the mutated region was amplified from Rd KW20 chromosomal DNA using primers F1 and MegaR (Table S2). The mega primer was then used in conjunction with primer mP2_R using pMSA16 as template to generate a linear repaired mutagenic product comprising of upstream flanking sequence, ompP2 lacking nucleotides encoding loop 2, kanamycin resistance cassette and downstream flanking sequence. This product was used to naturally transform H. influenzae Rd KW20 using a previously described method for the natural transformation of N. meningitidis [46]. The ompP2 loop 2 deletion in the resulting mutant (RdOmpP2 D91-99 ) was confirmed by PCR analysis and DNA sequencing. Growth curve assays carried out using liquid cultures showed no significant differences between RdOmpP2 D91-99 and the wild-type strain (data not shown).

Mutation of OmpP2 in H. influenzae
796-bp of ompP2 was removed from pMSA15 by inverse PCR using primers DP2F and P2DL2I_R. These primers incorporated BglII-sites allowing re-ligation of the amplicon to form pMSA17 (Table S1). This PCR template was then used in conjunction with primers F1 and P2FR to yield a linear mutagenic PCR product that was used to naturally transform H. influenzae Rd KW20 using a previously described method [46]. The ompP2 deletion in the resulting mutant (RdDompP2) was confirmed by PCR analysis and DNA sequencing. Growth curve assays carried out using liquid cultures showed no significant differences between RdDompP2 and the parental strain (data not shown).

Cell fractionation and confocal microscopy
Cell fractionation was undertaken as described previously [48]. For confocal microscopy, growth from overnight cultures was harvested by centrifugation (40006 g for 10 min), resuspended in 16 PBS, and the OD 600 adjusted to 1.0. 50 ml aliquots were then added to Knittel adhesive glass slides (SLS) and left at room temperature for 1 h. Bacteria were fixed using 4% paraformaldehyde for 10 min. Slides were washed four times in 16 PBS and then blocked with 1% BSA/PBS for 1 h. Surface-accessible PorA was labeled with mouse anti-meningococcal serosubtype P1.7 monoclonal primary antibody (1:25 diluted; NIBSC code: 01/514) for 1 h. Following washing in 16 PBS, goat anti-mouse IgGcascade blue conjugate (1:800 diluted; Invitrogen) was added for 1 h. Following further washes in 16 PBS, membranes were stained with FM1-43 (10 mM; Invitrogen) in Hank's Balanced Salt Solution (Invitrogen) for 30 min in the dark. Following a final 16 PBS wash, coverslips were mounted with Prolong Gold anti-fade (Invitrogen) and images acquired on a Zeiss LSM700 confocal microscope using a 636 objective. Images are single sections (300 nm) and data was collected from different fluorophores in separate channels. Images were processed using Image J and Adobe Photoshop.