Structural and functional insights into the role of BamD and BamE within the β-barrel assembly machinery in Neisseria gonorrhoeae

The β-barrel assembly machinery (BAM) is a conserved multicomponent protein complex responsible for the biogenesis of β-barrel outer membrane proteins (OMPs) in Gram-negative bacteria. Given its role in the production of OMPs for survival and pathogenesis, BAM represents an attractive target for the development of therapeutic interventions, including drugs and vaccines against multidrug-resistant bacteria such as Neisseria gonorrhoeae. The first structure of BamA, the central component of BAM, was from N. gonorrhoeae, the etiological agent of the sexually transmitted disease gonorrhea. To aid in pharmaceutical targeting of BAM, we expanded our studies to BamD and BamE within BAM of this clinically relevant human pathogen. We found that the presence of BamD, but not BamE, is essential for gonococcal viability. However, BamE, but not BamD, was cell-surface–displayed under native conditions; however, in the absence of BamE, BamD indeed becomes surface-exposed. Loss of BamE altered cell envelope composition, leading to slower growth and an increase in both antibiotic susceptibility and formation of membrane vesicles containing greater amounts of vaccine antigens. Both BamD and BamE are expressed in diverse gonococcal isolates, under host-relevant conditions, and throughout different phases of growth. The solved structures of Neisseria BamD and BamE share overall folds with Escherichia coli proteins but contain differences that may be important for function. Together, these studies highlight that, although BAM is conserved across Gram-negative bacteria, structural and functional differences do exist across species, which may be leveraged in the development of species-specific therapeutics in the effort to combat multidrug resistance.

Gram-negative bacteria, mitochondria, and plastids, such as chloroplasts, contain integral ␤-barrel outer membrane proteins (OMPs) 4 that play a myriad of pivotal physiological and structural functions, including nutrient acquisition, secretion, signal transduction, outer membrane biogenesis, and motility (1,2). In pathogenic bacteria, OMPs additionally assist in virulence by facilitating host colonization and exploiting immune responses as well as drug extrusion (3)(4)(5). Therefore, understanding the mechanisms that direct targeting and folding of OMPs is critical for development of pharmaceutical interventions to combat clinically important pathogens, including the recent emergence of multidrug-resistant Neisseria gonorrhoeae, the etiologic agent of gonorrhea. This sexually transmitted infection remains a major public health problem globally, and 78 million cases were estimated in 2012 (6). Recent quantitative proteomic investigations of OMPs in N. gonorrhoeae yielded new insights into cell envelope composition and identified new vaccine/drug protein targets that include LptD, TamA, TamB, and the BAM complex as well as a plethora of uncharacterized lipoproteins (7,8).
OMPs are first synthesized in the cytoplasm with an N-terminal leader sequence that routes them across the inner membrane into the periplasm by the Sec system (4,9,10). Periplasmic chaperones, such as SurA, FkpA, and/or Skp, then bind the nascent OMPs and escort them to the outer membrane where the ␤-barrel assembly machinery (BAM) then folds and/or inserts them into the outer membrane (3-5, 9, 10). BAM has been primarily investigated in Escherichia coli and Neisseria meningitidis (3,4,11,12). In E. coli, BAM is a five-protein com-plex consisting of BamA, an OMP itself, and four accessory lipoproteins, BamB, BamC, BamD, and BamE (3,5,11). In contrast, no BamB ortholog is present in Neisseria genomes, whereas RmpM has been identified as an additional accessory protein (11). BamA is the central component of BAM, and its removal results in loss of viability. Similarly, a bamD deletion causes lethality in E. coli and N. meningitidis, whereas lack of other accessory Bam lipoproteins results in various degrees of growth phenotypes and more subtle effects on cell envelope composition and integrity (3,11,(13)(14)(15)(16). In vitro, all lipoproteins are required for fully efficient folding of OmpT, a model OMP substrate of BAM (17,18). BamB-E are found within the periplasm anchored to the periplasmic leaflet of the outer membrane via a lipid moiety at their N terminus. Albeit still controversial, particularly with the recently reported structures of BAM (19 -22), the helix-grip domains of BamC have also been shown to be surface-exposed (23).
Over the past decade, the individual structures of all the Bam proteins have been reported from E. coli, and more recently, several groups reported the structure of fully assembled E. coli BAM (20 -22, 24 -39). These structures have provided molecular details about how the individual Bam proteins interact with one another and revealed that the barrel domain of BamA undergoes a large conformational change within the membrane not previously observed in OMPs (19 -22, 30, 38 -40). BamB and BamD were found to interact directly with BamA with BamD also interacting with the barrel domain. BamE was found to not only interact with BamD as shown previously (14) but also with BamA, bridging an additional interaction of BamA with BamD. These structures have contributed significantly toward our understanding of the architecture and dynamics of BAM; however, exactly how BAM functions in E. coli remains unknown. In Neisseria, only the structure of full-length BamA (N. gonorrhoeae) has been reported for BAM (30) along with the structure of RmpM (N. meningitidis) (41).
Growing lines of evidence build an appreciation for the existence of significant differences in homologous protein function, structure, and localization, often despite a close relatedness of the organisms. For instance, factor H-binding protein, which is incorporated into the BEXSERO meningococcal B vaccine, is a surface-localized protein in N. meningitidis but not in N. gonorrhoeae (42). Furthermore, the protein responsible for transporting lipopolysaccharide in E. coli and lipooligosaccharide in Neisseria to the cell surface, LptD, is non-essential in N. meningitidis but essential in E. coli and N. gonorrhoeae (7,43,44).
To gain additional insight into the role of BamD and BamE within BAM in Neisseria and to aid in future therapeutic development, here we have performed mutagenesis and knockout studies to assay the effects on Neisseria growth and OMP assembly. Our work shows that BamD, but not BamE, is essential for viability and that BamE, but not BamD, is surface-exposed, similar to what has been observed for BamC in E. coli. However, in the absence of BamE, BamD did become surfaceexposed with a concomitant increase in antibiotic susceptibility and production of membrane vesicles with altered OMP composition, providing further evidence that BamE may be a new vaccine target against N. gonorrhoeae. Furthermore, to better guide future investigations on Neisseria BAM and to assist in structure-based therapeutic methods, we have determined the X-ray crystal structures of both BamD and BamE from N. gonorrhoeae. These studies show that Neisseria BamD and BamE share overall folds with their E. coli orthologs, but there are differences that may be functionally important.  (Fig. S2).

BamD GC is an essential BAM component in N. gonorrhoeae
In E. coli and N. meningitidis, both BamA and BamD are essential proteins, and their depletion results in OMP folding, stability, and assembly defects (11, 46 -48). Recently, we demonstrated the essential nature of BamA GC in N. gonorrhoeae (8), but bamD GC was designated dispensable due to the successful generation of a transposon mutant (45). This mutant, however, showed pleiotropic phenotypes, including smaller cell size, cratered and crinkled colony morphology, and reduced transformation competence.
Based on studies in E. coli and N. meningitidis (11,46), we reasoned that BamD GC is also essential for N. gonorrhoeae viability. To test this hypothesis, we first constructed a strain in the N. gonorrhoeae FA1090 background that carried an additional copy of bamD GC placed under an isopropyl ␤-D-1-thiogalactopyranoside (IPTG)-inducible promoter in a different site on the chromosome. We next insertionally inactivated the bam-D GC gene in its chromosomal locus using the kanamycin resistance cassette. The resulting strain, ⌬bamD GC /P lac ::bamD GC , formed robust colonies in the presence of the inducer but failed to grow when subcultured on solid medium lacking IPTG (Fig.  1A). To deplete BamD during growth in liquid medium, we applied an experimental strategy utilized previously to diminish levels of the N. gonorrhoeae essential proteins BamA, Obg, and GmhA (8,49,50). After harvesting from solid medium supplemented with IPTG, bamD::kan/P lac ::bamD was transferred to broth with or without IPTG. Following 3-h incubation, both cultures were adjusted to the same optical density (shown as time 0 h on the graph in Fig. 1B) and cultured under permissive and non-permissive conditions for another 6 h. Omitting IPTG in the liquid medium prevented bacterial growth (Fig. 1B), which was concomitant with BamD GC depletion as shown by immunoblotting with anti-BamD GC antiserum (Fig. 1C). Together, these studies confirmed the essential nature of BamD in N. gonorrhoeae.

N. gonorrhoeae lacking BamE GC displays a slower growth rate in liquid medium under standard conditions
To characterize the role of BamE GC , we cloned and purified a construct lacking the predicted lipoprotein signal peptide and tagged with a C-terminal His 6 tag and used the purified pro-tein to raise polyclonal rabbit antiserum. The purified protein migrated in SDS-PAGE according to a predicted molecular mass of ϳ12 kDa, corresponding to the recombinant protein.
Subsequently, we created a null bamE GC mutant, ⌬bamE GC , and a complemented strain ⌬bamE GC /P lac ::bamE GC , in N. gonorrhoeae FA1090. The anti-BamE GC antiserum cross-reacted with wildtype whole-cell lysates. As expected, no signal was detected in the bamE GC knockout strain, whereas the complemented strain expressed BamE GC at levels proportional to the concentrations of IPTG added ( Fig. 2A). The use of 20 M IPTG resulted in amounts of BamE GC closely resembling the native A, cells of N. gonorrhoeae FA1090 wildtype, isogenic ⌬bamE GC , and ⌬bamE GC /P lac ::bamE GC were harvested from solid medium supplemented with IPTG (as indicated) and subjected to SDS-PAGE followed by immunoblotting with anti-BamE GC antiserum. Migration of a molecular mass marker (kDa) is indicated on the left. B and C, non-piliated colonies of FA1090 wildtype, isogenic ⌬bamE GC , and ⌬bamE GC /P lac ::bamE GC were suspended in liquid medium with IPTG to an A 600 of 0.1 and incubated under standard aerobic conditions for 3 h. Then, cultures were either back-diluted in fresh medium and cultured for an additional 6 h, and growth was monitored at

Structure and function of BamE and BamD
protein pool in the wildtype strain and consequently was chosen in further complementation experiments. The colony size of ⌬bamE GC was similar to that in a parental strain, but a slower proliferation rate was observed during midlogarithmic growth in liquid medium. At the end of the experiment, however, the mutant culture reached the same density as the wildtype, and the apparent lag in growth was fully rescued in ⌬bamE GC /P lac ::bamE GC (51) (Fig. 2B). Culture supernatants derived from ⌬bamE GC contained increased amounts of cytoplasmic protein markers such as Zwf, Obg, and GmhA, suggesting increased cell lysis (51). Interestingly, culturing the mutant in chemically defined Graver-Wade medium significantly reduced this phenotype. Slower growth in liquid medium was also observed in the Caulobacter crescentus ⌬bamE but not in E. coli, N. meningitidis, Pseudomonas aeruginosa, or Salmonella enterica serovar Typhimurium (14,(52)(53)(54). In contrast, no significant fitness differences, as measured by counts of colony-forming units (cfu), were noted when the wildtype, ⌬bamE GC , and ⌬bamE GC /P lac ::bamE GC were maintained on solid medium under standard aerobic conditions as well as conditions mimicking different microecological niches in the human host, including iron deprivation, presence of normal human serum, and anaerobiosis (Fig. 2C).

Expression of BamE GC and BamD GC
Limited information is available regarding expression of BAM components as most research efforts have focused on understanding the architecture and protein interactions of this protein complex. Therefore, to further characterize the accessory lipoproteins BamD GC and BamE GC , we examined their expression patterns in the wildtype FA1090 throughout growth in liquid medium, during exposure to environmental stimuli relevant to different infections sites in the human host, and in a panel of 36 different gonococcal isolates.
Both BamD GC and BamE GC were continuously expressed throughout all stages of N. gonorrhoeae growth (Figs. 1A and 3A). A similar expression pattern was reported for BamA GC (8). Moreover, although levels of BamD GC remained unchanged upon exposure of N. gonorrhoeae to aerobic and anaerobic conditions, upon iron limitation and in the presence of normal human serum, expression of BamE GC was noticeably induced during anaerobiosis (Fig. 3B). This could suggest an additional requirement for BamE GC during anoxia, but this potential role seems not to be sufficiently significant as fitness of N. gonorrhoeae lacking bamE GC was not affected under these conditions (Fig. 2C).
Analysis of SNPs showed high conservation of both BamE GC and BamD GC among different N. gonorrhoeae isolates (Fig. S2). Corroborating this observation, the anti-BamE GC and anti-BamD GC antiserum cross-reacted with whole-cell lysates derived from common laboratory strains (FA1090, F62, MS11, FA19, and 1291) and temporally and geographically diversified clinical isolates, including the 2016 World Health Organization (WHO) reference strains (Fig. 3C). In addition, these experiments demonstrated that, similarly to BamA GC (8), the accessory lipoproteins BamE GC and BamD GC are ubiquitously expressed in a highly diverse pool of gonococcal isolates, fur-ther underscoring the potential of BAM as a target for new therapeutic interventions.

Subcellular localization studies of BamD GC and BamE GC
Using a quantitative proteomic approach, we have previously identified BamA GC , BamD GC , and BamE GC in both cell envelopes and naturally released membrane vesicles derived from four different N. gonorrhoeae strains (7,55). Intriguingly, differences in homologous proteins' subcellular localization between even closely related organisms such as N. meningitidis and N. gonorrhoeae have been reported (42). Furthermore, recent findings showed that E. coli BamC is exposed on the surface and accessible to antibodies and proteases, which challenges the dogma for the architecture of BAM (3,23). Therefore, we first confirmed the outer membrane localization and surface exposure of BamA GC (8). Subsequently, we sought to examine the subcellular location of BamD GC and BamE GC . Subproteome fractions isolated from wildtype N. gonorrhoeae FA1090 were separated by SDS-PAGE and probed with antisera against BAM LGB1 LG14 LG2 LGB26 LG20 , and under anaerobiosis (ϪO 2 )) were assessed by probing the whole-cell lysates with respective antibodies. C, 37 strains of N. gonorrhoeae, as indicated above the immunoblots, were grown concurrently on solid medium for 20 h in 5% CO 2 at 37°C, and bacteria were collected, lysed, and processed for immunoblotting. In all experiments, samples containing the whole-cell lysates were matched by equivalent A 600 units, resolved in a 4 -20% Tris-glycine gel, and transferred onto nitrocellulose. Immunoblot analysis was performed using polyclonal rabbit antisera against BamD GC and BamE GC . Migration of a molecular mass marker (kDa) is indicated on the left.

Structure and function of BamE and BamD
proteins as well as a cytoplasmic protein marker, Obg GC . As expected, BamA GC and both lipoproteins, BamE GC and Bam-D GC , were detected in the cell envelope and membrane vesicle fractions but not in the cytoplasm or culture supernatants, whereas Obg GC was primarily localized to the cytoplasmic compartment (Fig. 4A). Furthermore, to assess the surface exposure of BamE GC and BamD GC , dot blotting and protease treatment experiments using whole cells were performed according to optimized protocols for N. gonorrhoeae that ensure intactness of the cells (8). Anti-BamE GC antiserum cross-reacted with intact wildtype and ⌬bamE GC /P lac ::bamE GC cells but not with cells of the ⌬bamE GC mutant. In contrast, BamD GC was not recognized on wildtype cells by anti-BamD GC antiserum unless the cells were lysed (Fig. 4B). Corroborating these findings, exposing intact gonococci to increasing concentrations of trypsin resulted in detection of decreased amounts of BamE GC , similar to the surface-exposed BamA GC , whereas levels of BamD GC ; a periplasmic marker, SurA; and Obg GC remained unchanged (Fig. 4C). These studies suggested that at least part of the cellular pool of BamE GC is localized on the outside of the cell, whereas BamD GC faces the periplasmic side of the outer membrane. In addition, in the absence of BamE GC , BamD GC became accessible to antibodies and susceptible to protease treatment. The levels of SurA remained unchanged in ⌬bamE GC in comparison with wildtype cells, excluding the possibility that surface localization of BamD GC in the ⌬bamE GC mutant is solely attributable to altered outer membrane integrity (Fig. 4C). It has been suggested that both BamC and BamE stabilize the BamA-BamD interaction. In E. coli, upon BamE depletion, BamA becomes dramatically susceptible to exogenously added protease, whereas periplasmic proteins, including SurA and BamD, are completely unaffected (56). In contrast, our studies suggest that loss of BamE GC weakens the BamA GC -BamD GC interaction, causing a surface exposure of BamD GC with no significant increase in the protease sensitivity of BamA GC . Furthermore, although the absence of BamE GC had no significant impact on the steady-state-levels of any other BAM components in the cell envelopes of E. coli (56) and N. gonorrhoeae (Fig. 5C), depletion of BamD GC resulted in an increase in the cellular pool of both BamA GC and BamE GC (Fig. 5C).

Lack of BamE GC affects the proteome of cell envelopes and membrane vesicles
In E. coli, loss of BamB, but not BamC or BamE, results in significant alterations in the barrier function of the outer membrane, manifested by elevated sensitivity to several antibiotics (56). In contrast, BamE-depleted cells of N. meningitidis, C. crescentus, and S. enterica are deficient in both outer membrane assembly and integrity (11,52,53). To examine the impact of BamE GC on membrane permeability, we used several different methods and growth media, including disc diffusion, E-tests, and phenotypic microarrays (51) (Tables 1 and 2). In disc diffusion assays, 14 different conditions were tested with a total of 10 compounds, including detergents and antibiotics (vancomycin, carbenicillin, and polymyxin B), significantly impacting the ⌬bamE GC strain in comparison with the wildtype (Table 1). Furthermore, E-tests demonstrated significantly lowered minimal inhibitory concentrations (MICs) for cefuroxime, azithromycin, ciprofloxacin, and polymyxin B ( Table  2). Previously applied phenotypic microarrays with 1,056 conditions and performed in defined liquid medium showed three and six conditions uniquely beneficial and detrimental, respectively, to the ⌬bamE GC mutant in comparison with six null mutants in novel gonorrhea vaccine candidates (51). The three beneficial compounds were osmolytes, whereas sodium benzoate and chromium chloride attenuated the growth of ⌬bamE GC . The mutant was also negatively affected by nalidixic acid, rifampicin, doxycycline, and cefsulodin, which all exert different mechanisms to kill bacteria. Similarly, loss of BamE A, wildtype N. gonorrhoeae FA1090 harvested during the midlogarithmic phase was subjected to proteome extraction to separate cytoplasmic proteins (C), cell envelopes (CE), naturally released membrane vesicles (MV), and soluble proteins in culture supernatants (SS). Subproteome fractions, normalized based on the total amount of protein, were resolved by SDS-PAGE and probed with polyclonal antisera against the indicated proteins. B, N. gonorrhoeae strains, as shown above the graphs, were cultured in liquid medium, harvested, and suspended to the same A 600 of 2.0. Intact as well as lysed cells were spotted onto nitrocellulose membranes and probed with polyclonal antisera against BamE GC , BamA GC , BamD GC , and Obg GC . C, N. gonorrhoeae FA1090 wildtype and isogenic ⌬bamE GC cultures at an A 600 of ϳ1.0 were harvested, suspended in sterile PBS, and incubated at 37°C for 1 h without or with increasing concentrations of trypsin as indicated above the immunoblot. The reaction was stopped with the addition of PMSF, cells were washed, and individual protein profiles were analyzed by immunoblotting with specific antisera against BamE GC , BamA GC , BamD GC , SurA GC , and Obg GC . Migration of a molecular mass marker (kDa) is indicated on the left.

Structure and function of BamE and BamD
caused increased sensitivity to nalidixic acid, rifampicin, and carbenicillin in C. crescentus (52).
The sensitivity phenotype of ⌬bamE GC suggested defects in membrane permeability, but examination of the general cell envelope protein profile did not show apparent alterations (Fig.  5A). We therefore analyzed naturally released membrane vesicles from wildtype and ⌬bamE GC . There was an increase in the abundance of several protein species as revealed by SDS-PAGE and Coomassie staining (Fig. 5A) and a 3-fold higher shedding of membrane vesicles from ⌬bamE GC in comparison with the wildtype (Fig. 5B). Outer membrane vesiculation is a well-recognized indicator of cell envelope stress (57). To gain further insights into the scale of OMP defects associated with loss of BamE GC , immunoblotting with antisera specific to 10 OMPs was performed on cell envelopes and membrane vesicles derived from the wildtype and ⌬bamE GC . These studies demonstrated elevated levels of seven OMPs within the cell envelope fraction with the most significant alterations observed for AniA, Laz, and Ng-MIP (Fig. 5C). This effect was exacerbated in the membrane vesicles where the amounts of AniA and Laz increased over 7-fold, NGO2139 (MetQ) increased 5-fold, and LptD increased almost 3-fold (Fig. 5D). In contrast, the other Bam components and MtrE remained unaltered, whereas TamA was about 2-fold depleted. These results suggest a specific contribution of BamE GC to cell envelope biogenesis.

The structures of N. gonorrhoeae BamD and BamE
To gain insights into the structure and function of BamD GC and BamE GC and to facilitate the future targeting of BAM with small molecule inhibitors, we obtained recombinant BamD GC and BamE GC for structural studies as described under "Experimental procedures." The structure of BamD GC was solved by molecular replacement to 2.5-Å resolution with final R/R free values of 0.24/0.29 and contained one molecule per asymmetric unit. The BamD GC structure closely resembles that from E. coli, consisting of five tetratricopeptide repeat (TPR) domains and having an overall root mean square deviation (r.m.s.d.) of 2.05 Å (residues 30 -257) (35,36,58) (Fig. 6A). The shape of BamD GC is slightly more bent than the E. coli ortholog, which is best observed when aligning both structures along TPR1 only (Fig.  6B). Upon closer inspection, the individual TPR domains are more conserved than the overall r.m.s.d. suggests with r.m.s.d. values for TPR1 alone of BamD GC calculated to be 1.07 Å (residues 30 -66), for TPR2 0.704 Å (residues 67-104), for TPR3 0.583 Å (residues 105-159), for TPR4 0.675 Å (residues 160 -210), and for TPR5 0.471 Å (residues 211-245) (Fig. 6C). Additionally, the conserved arginine residue of BamD GC (Arg-200), important for forming a salt bridge interaction with a conserved glutamate residue of BamA (Glu-373) (19,21,22,31,59), was also found to be well-conserved and perfectly positioned to serve the same role in Neisseria (Fig. 6D).
The BamE GC structure was solved by selenium single-wavelength anomalous diffraction to 2.45-Å resolution with final R/R free values of 0.20/0.24 and contained two molecules per asymmetric unit with each monomer interacting with the other through a ␤-␤ interaction along residues 30 -36 (Fig. 7A). Each monomer contained the core ␣␣␤␤␤ fold found in other reported BamE structures with a calculated r.m.s.d. between chain A and chain B of 0.61 Å (Fig. 7B) (Fig. 7, C  and D). Interestingly, BamE GC contains an additional C-terminal helix not observed in the other reported BamE structures.

Discussion
BAM is an essential multicomponent complex that resides in the outer membranes of Gram-negative bacteria, making it an attractive target for engineering a novel class of antibiotics that

Structure and function of BamE and BamD
do not have to cross cellular membranes. Recent studies show that the composition of BAM varies by bacterial species with E. coli having at least five components within the core complex, N. meningitidis and N. gonorrhoeae having four BAM proteins (lacking a BamB ortholog) and RmpM, and C. crescentus having BamF instead of BamC (11,41,60). The structures of all the components of BAM in E. coli have been solved, including the full complex (19 -22). We have previously reported the structure of the central and essential component of BAM, BamA, from N. gonorrhoeae, and here we report the structures of BamD GC and BamE GC (Figs. 6 and 7). The structure of BamD GC most closely matches that of the E. coli ortholog; however, there are small structural differences. The structure of Neisseria BamE GC , however, deviates more significantly compared with the E. coli ortholog with an additional C-terminal helix not previously observed in BamE. The possible role of this helix will be explored in future studies to determine whether it is important for BAM function.
In addition to the structures, we studied the roles of BamD GC and BamE GC within BAM in N. gonorrhoeae. We found that BamD GC is essential for viability, whereas BamE GC was dispensable, aligning well with what has been observed previously for N. meningitidis (11) and E. coli (14,46,47). Analogous to the presence of BamC on the surface of E. coli (23), we found that BamE GC , but not BamD, was surface-displayed in our surface labeling and protease shaving experiments (Figs. 4, B and C, and 8A). Intriguingly, we also observed BamD GC on the surface of gonococci but only in BamE GC -depleted cells (Figs. 4, B and C, and 8B). This suggests that the absence of BamE GC leads to destabilization of BAM, possibly weakening the interaction between BamA GC and BamD GC and resulting in surface exposure of BamD GC . There was, however, no significant increase in the protease sensitivity of BamA GC . It is possible that BamC GC and (an)other yet unrecognized protein(s) docked into the BAM complex influence BamA GC conformation. In E. coli, both BamE and BamC are important for stabilizing the BamA-BamD interaction, and BamE may also modulate the conformational state of BamA (56).
The lack of BamE GC was also accompanied by an increase in antibiotic susceptibility (Tables 1 and 2 and Ref. 51) and a significantly greater release of membrane vesicles containing altered levels of new vaccine antigens (Fig. 8B). Cumulatively, our experiments confirmed that BamD GC and BamE GC localized to the cell envelope and membrane vesicles; showed the ubiquitous expression of BamA GC (8), BamD GC , and BamE GC in a diverse pool of gonococcal isolates, further underscoring the potential of BAM as a target for novel antibiotics and vaccines against Neisseria; and, importantly, revealed additional interspecies differences existing within BAM, illuminating the need for parallel studies in different organisms to enhance our understanding of cell envelope biogenesis. Together, these studies indicate that, although BAM is conserved across all Gram-negative bacteria, structural and functional differences do exist across bacterial species and may be utilized in the development of species-specific antibiotics and vaccines in the effort to combat multidrug resistance.

Structure and function of BamE and BamD
ated colonies were used for transformation, whereas non-piliated variants were used in all other experiments. E. coli strains were grown either on Luria-Bertani agar (LBA; Difco) or cultured in Luria-Bertani broth (LB; Difco) at 37°C. Antibiotics were used in the following concentrations: for N. gonorrhoeae, kanamycin, 40 g/ml; and erythromycin, 0.5 g/ml; for E. coli, kanamycin, 50 g/ml; erythromycin, 250 g/ml; and carbenicillin, 50 or 100 g/ml as specified in the text.

Genetic manipulations
Oligonucleotides used in this study (Table S1) were designed using SnapGene software version 2.8 (GSL Biotech LLC) based on the genomic sequence of N. gonorrhoeae FA1090 (NC_002946). Primers were synthesized by Integrated DNA Technologies. Genomic DNA of N. gonorrhoeae FA1090 was purified with the Wizard Genomic DNA Purification kit (Promega) or purchased directly from ATCC and used as template in PCRs with applicable oligonucleotides and Q5 high-fidelity DNA polymerase (New England Biolabs). PCR products and plasmid DNA were purified using a QIAprep Spin Miniprep kit (Qiagen). Obtained genetic constructs were verified by Sanger sequencing at the Center for Genomic Research and Biocomputing at Oregon State University and USA Macrogen. Transformation of N. gonorrhoeae was performed as described previously (55).
The N. gonorrhoeae FA1090 conditional bamD knockout, bamD::kan/P lac ::bamD, was constructed according to the following steps. First, an additional copy of the bamD gene (NGO0277) under lac regulatory sequences, P lac ::bamD (8), was placed at an unlinked chromosomal locus between the lctP and aspC genes using the Neisseria Insertional Complementation System, pGCC4 (65), to yield the FA1090 P lac ::bamD strain. Subsequently, a 536-bp DNA fragment containing the N-terminal part of the bamD gene and upstream DNA region was amplified with primers BamD-Up-F and BamD-Up-R, digested with EcoRI/KpnI, and introduced into similarly treated pUC18K (66), yielding pUC18K-BamD-Up. Next, the downstream DNA fragment for allelic replacement (562 bp) was amplified with primers BamD-Down-F/BamD-Down-R, digested with BamHI/HindIII, and ligated into BamHI/ HindIII-cleaved pUC18K-BamD-Up. The pUC18K-⌬bamD::kan was linearized with HindIII and introduced into FA1090 P lac ::bamD. Transformants were selected on GCB supplemented with kanamycin and 0.05 mM IPTG and verified for disruption of the bamD gene with the kanamycin resistance cassette by PCR with primers BamD-Ver-F/BamD-Ver-R and immunoblotting analysis using BamD antiserum (8).
To generate a clean deletion of bamE in N. gonorrhoeae FA1090, the upstream region of NGO1780 was amplified with  B B B B B B B B B B B B B B  A, under wildtype BAM conditions, BamE GC , but not BamD GC , was found localized at the cell surface of gonococci. B, in the absence of BamE GC , topological changes in BAM occur, including BamD GC becoming surface-exposed, which is also accompanied by an elevation in the release of membrane vesicles with altered OMP composition.

Structure and function of BamE and BamD
primers BamE-Up-F/BamE-Up-R. The 757-bp product was cleaved with SacI/KpnI and cloned into similarly treated pUC18K, yielding pUC18K-BamE-Up. Subsequently, the downstream region from the gene encoding BamE was amplified with primers BamE-Down-F/BamE-Down-R, and the obtained 724-bp product was cloned into BamHI/HindIII-treated pUC18K-BamE-Up. The final product, pUC18K-⌬bamE, was used for an allelic exchange of bamE with the kanamycin resistance cassette as described above. Deletion of bamE was confirmed by PCR with primers BamE-Ver-F/BamE-Ver-R using chromosomal DNA isolated from wildtype FA1090 as controls and by probing the whole-cell lysates of wildtype and ⌬bamE with antiserum against BamE.
To complement the N. gonorrhoeae FA1090 ⌬bamE mutant, first the P lac ::bamE construct was generated by amplification of bamE with native ribosome-binding site using primers cBamE-F/cBamE-R. The 41-bp PCR product was digested with FseI and inserted into ScaI/FseI-cleaved pGCC4, yielding pGCC4-BamE. Next, pGCC4-BamE was introduced into the ⌬bamE mutant by transformation, and clones were selected on GCB containing erythromycin and validated by PCR with primers pGCC4-Ver-F/pGCC4-Rev-R as well as immunoblotting with anti-BamE antiserum.
To obtain pET28-rBamE used for production of recombinant BamE (rBamE), which lacked the native signal peptide and contained a C-terminal His 6 tag, the ngo1780 gene was amplified with primers rBamE-F/rBamE-R and cloned into NcoI/ HindIII-digested pET28a.
The gene encoding SurA (ngo1714) lacking the DNA encoding signal peptide was amplified using primers SurA-F/SurA-R. The subsequent PCR product was digested with NcoI/HindIII and ligated into similarly cut pRSF-NT to create a TEV protease-cleavable C-terminal His 6 -tagged fusion. For structural studies, the BamD and BamE coding regions, starting after the N-terminal cysteine, were amplified from N. gonorrhoeae strain FA1090 genomic DNA (ATCC) and subcloned into the pHIS-parallel2 vector using NcoI and XhoI restriction sites. All sequences were verified by sequencing analysis (primers available upon request).

Growth assays
Depletion of BamD was achieved by applying an experimental strategy utilized previously to diminish levels of N. gonorrhoeae BamA, Obg, and GmhA (8,49,50). N. gonorrhoeae FA1090 bamD::kan/P lac ::bamD was harvested from GCB with 0.05 mM IPTG (permissive conditions), and the A 600 was adjusted to 0.1. Bacteria were washed twice and cultured in GCBL with or without IPTG at a final concentration of 0.05 mM. After 3 h, both cultures were diluted to an A 600 of 0.1 and cultured in fresh GCBL with or without IPTG for another 6 h. At every hour, A 600 measurements were taken, and samples were withdrawn for immunoblotting analysis. Three biological replicates of the experiment were performed. Mean values and corresponding S.D. are reported.
The growth kinetics of FA1090 wildtype, ⌬bamE, and ⌬bamE/P lac ::bamE were performed in GCBL under standard growth conditions. Bacteria were collected from GCB and suspended in GCBL to an A 600 of 0.1. Media for growing ⌬bamE/P lac ::bamE were supplemented with 0.02 mM IPTG. Following 3 h of incubation at 37°C with aeration (220 rpm), bacterial cultures were back-diluted to an A 600 of 0.1 in fresh GCBL and cultured for an additional 6 h. Samples were withdrawn for A 600 measurements every hour (n ϭ 3; mean Ϯ S.D.).
To assess the viability of N. gonorrhoeae lacking bamE during host-relevant in vitro growth conditions, colonies of FA1090 wildtype, ⌬bamE, and ⌬bamE/P lac ::bamE were collected from GCB, suspended in GCBL to an A 600 of 0.1, and cultured for 3 h at 37°C with aeration. Subsequently, the cultures were normalized to an A 600 of 0.2, serially diluted, and plated on solid medium for standard growth conditions (GCB), iron-limiting conditions, NHS, and anaerobic conditions as described above. All media were additionally supplemented with 0.02 mM IPTG.
The cfu values were scored after 22 and 48 h for aerobic and anaerobic conditions, respectively. Experiments were performed on three separate occasions, and mean cfu values with corresponding S.D. are reported.

Antimicrobial susceptibility testing
Antimicrobial susceptibility was tested using a slightly modified Kirby-Bauer (disk diffusion) method (67) and E-test. In disc diffusion experiments, FA1090 wildtype, isogenic ⌬bamE, and ⌬bamE/P lac ::bamE strains were collected from GCB, and the suspensions were adjusted in GCBL to match an A 600 of 0.2. Cell suspensions (100 l) were immediately plated on GCB, and 6-mm filter paper disks (Whatman) impregnated with 10 l of tested compounds, as indicated below, were placed on the surface of the agar. The zones of inhibition in mm were measured after 22 h. Experiments were performed in biological triplicates, and mean values with S.D. are presented.
MICs for cefuroxime, cefotaxime, azithromycin, tetracycline, ciprofloxacin, polymyxin B, ampicillin, and benzylpenicillin were determined using an E-test (Biomerieux) according to the manufacturer's recommendations. Each determination was performed on three separate occasions using fresh bacterial cultures, and the consensus MIC obtained in at least two trials was reported.

Protein localization assays
Subcellular fractionations, immunodotting, and trypsin accessibility studies were performed following procedures described previously (8). Briefly, N. gonorrhoeae FA1090 wildtype and ⌬bamE at the midlogarithmic phase of growth were used to extract proteins from the cytosolic, cell envelope, membrane vesicle, and soluble supernatant fractions. Cell envelopes were separated from cytoplasmic proteins by a sodium carbonate extraction method and differential centrifugation, whereas culture supernatants were subjected to filtration and ultracentrifugation to separate naturally released membrane vesicles from soluble proteins. Quantification of membrane vesicles was achieved by calculating the protein content in the membrane vesicles to 1.0 liter of original culture volume/OD unit (mg liter Ϫ1 OD unit Ϫ1 ) as described (68).
In immunodotting and protease susceptibility studies, intact bacterial cells were used (8). For immunodotting, bacteria were suspended in GCBL to an A 600 of 0.1, cultured with aeration for 3 h, harvested, and spotted as 5-l suspensions onto a nitrocel-

Structure and function of BamE and BamD
lulose membrane after adjusting the A 600 to 2.0. The samples were dried at room temperature for 15 min and subjected to immunoblotting.
In trypsin shaving assays, gonococci were subcultured in GCBL for 3 h after collecting from solid medium, diluted to an A 600 of 0.1, and cultured until an A 600 of ϳ1.0 was reached. Bacteria were gently harvested and suspended in sterile PBS, pH 8.0, to an A 600 of 2.5, and 500-l suspensions were incubated for 1 h at 37°C with trypsin at final concentrations of 0, 40, or 80 g/ml. To deactivate trypsin, 10 l of 50 mM phenylmethylsulfonyl fluoride (PMSF) was added, bacteria were washed with GCBL and subjected to SDS-PAGE, and trypsin accessibility of selected proteins was detected by immunoblotting with polyclonal antiserum.

Purification of rBamE and rSurA and preparation of polyclonal antisera
An overnight culture of E. coli BL21(DE3) carrying either pET28a-rBamE or pET28a-rSurA was back-diluted into 3.0 or 1.0 liter of LB broth, respectively, supplemented with kanamycin and incubated with aeration at 37°C. The production of rBamE and rSurA was induced with 0.1 and 1 mM IPTG, respectively, during midlogarithmic growth. Bacterial cells were collected by centrifugation 3 h after induction, suspended in lysis buffer (500 mM NaCl, 10 mM imidazole, 20 mM Tris-HCl, pH 8.0, and Complete EDTA-free protease inhibitor tablet (Roche Applied Science)), and lysed by passaging through a French pressure cell at 12,000 p.s.i. Unbroken cells and cell debris were removed by centrifugation at 16,000 ϫ g for 30 min at 4°C. The cell-free lysate was passed through a 0.22-m filter unit (VWR International) and applied onto Bio-Scale Mini Profinity immobilized metal affinity chromatography cartridges (Bio-Rad). Loosely bound proteins were removed with 10 column volumes of wash buffer (20 mM Tris, pH 8.0, 500 mM NaCl, 40 mM imidazole), and proteins were eluted with a 40 -250 mM imidazole gradient using an NGC Purification System (Bio-Rad). Fractions containing rBamE were pooled and dialyzed against 20 mM Tris, pH 8.0, 10% glycerol, whereas fractions containing eluted rSurA were incubated overnight at 4°C with TEV protease in a 1:20 ratio to remove the His 6 tag. After concentrating the sample to 5 ml using Vivaspin 20 centrifuge concentrators (GE Healthcare), proteins were subjected to size exclusion chromatography using a HiLoad 16/600 Superdex 75 prep grade column (GE HealthCare) with phosphatebuffered saline (PBS) as running buffer. Finally, fractions containing rSurA were concentrated using a Vivaspin 20 centrifuge concentrator.
Polyclonal antisera against purified rSurA and rBamE were prepared by Pacific Immunology Corp. using a 13-week antibody production protocol and two New Zealand White rabbits under Animal Protocol 1 approved by the Institutional Animal Care and Use Committee and the National Institutes of Health Animal Welfare Assurance Program (A4182-01) in a certified animal facility (United States Department of Agriculture 93-R-283). The rabbit polyclonal anti-BamD antiserum were obtained and evaluated previously (8).

Expression and purification of recombinant N. gonorrhoeae BamD and BamE
For expression of BamD GC and BamE GC for structural studies, each construct was introduced into E. coli BL21(DE3) chemically competent cells, then plated onto LBA supplemented with carbenicillin (100 g/ml), and incubated overnight at 37°C. For native expression, a single colony was used to inoculate a 5-ml LB with carbenicillin (50 g/ml) starter culture, which was cultured to an A 600 of ϳ1.0. The cells were then washed with fresh LB and inoculated into 2 liters of 2ϫ YT (16 g/liter Tryptone, 10 g/liter yeast extract, 5.0 g/liter NaCl) medium supplemented with carbenicillin (50 g/ml). Bacteria were cultured at 37°C until an A 600 of ϳ0.8 was reached, expression was induced with 0.2 mM IPTG, and bacteria were grown an additional 8 h at 25°C before harvesting.
For selenomethionine-substituted BamE GC , E. coli B834(DE3) cells were transformed and plated onto LBA overnight. A single colony was used to inoculate a 25-ml LB with carbenicillin (50 g/ml) culture and allowed to grow at 37°C to an A 600 of 0.8 -1.0. The cells were then centrifuged, washed three times with minimal medium lacking methionine, and resuspended in 6 ml of wash medium, and then 1 ml was added to six flasks containing 1 liter each of minimal medium supplemented with selenomethionine (40 mg/liter) and carbenicillin (50 g/ml). These cultures were grown at 37°C until the A 600 was between 0.6 and 0.8, then induced with 0.5 mM IPTG, and allowed to grow an additional 24 h before harvesting.
For protein purification, cells were resuspended in PBS (12 mM phosphate buffer, pH 7.4, 137 mM NaCl, 2.7 mM KCl) at 5 ml/g of cell paste and supplemented with DNase I (10 g/ml) and PMSF (0.2 mM). The cells were then lysed by two passages through an Avestin C3 Emulsiflex, and the lysates were centrifuged at 39,000 ϫ g for 45 min. The supernatants were then applied to a pre-equilibrated 5-ml nickel-nitrilotriacetic acid resin column. The column was washed with at least 5 column volumes of PBS, a pre-elution was performed using PBS with 25 mM imidazole, and a final elution was done with PBS containing 250 mM imidazole. To remove the N-terminal His tag, TEV protease was added to the final protein sample along with 1 mM DTT and 0.5 mM EDTA and dialyzed overnight at 4°C in PBS. The dialyzed samples were then again applied to a 2-ml preequilibrated nickel-nitrilotriacetic acid resin column. The filtrate was collected, concentrated, and then applied to a HiPrep Sephacryl S-100 HR gel filtration column (GE Healthcare) in PBS as a final purification step. Peak fractions were confirmed by SDS-PAGE analysis, pooled, and concentrated to 10 mg/ml.

Crystallization and structure determination of BamD and BamE from N. gonorrhoeae
For crystallization, the samples were screened using commercial sparse-matrix crystallization screens on a TTP LabTech Mosquito Crystal crystallization robot using the hanging-drop vapor-diffusion method with a drop ratio of 1:1 (protein:well solution), and lead conditions were further optimized. BamD GC was crystallized in final conditions of 1.0 M LiCl, 100 mM HEPES, pH 7.0, 10% PEG 6000. Native and selenomethionine-substituted BamE GC crystals were grown in

Structure and function of BamE and BamD
final conditions of 10 mM zinc sulfate heptahydrate, 100 mM MES monohydrate, pH 6.5, 25% PEG monomethyl ether 550. Crystals were harvested directly from their drops, cryoprotected in mother liquor solution containing 20% glycerol, and flash frozen in liquid nitrogen until data collection.
Data sets for native BamD GC and BamE GC were collected at the GM/CA-CAT beamline 23ID at the Advanced Photon Source, Argonne National Laboratory. Data sets for selenomethionine-substituted BamE GC were collected at beamline X25 at the National Synchrotron Light Source, Brookhaven National Laboratory. All data were processed using HKL2000 (69) or xia2 (70). The BamD GC structure was solved by molecular replacement with PHASER within PHENIX (71) in space group P4 3 2 1 2 to 2.5-Å resolution using the E. coli BamD structure (Protein Data Bank code 2YHC) as a search model. The BamE GC structure was solved by selenium single-wavelength anomalous diffraction in space group P2 1 2 1 2 to 2.45-Å resolution with AutoSol within PHENIX (71). A single selenium site per monomer was used to calculate initial phases followed by AutoBuild (71). All manual model building was performed with Coot (72), and refinement was performed with phenix.refine within PHENIX (71). Structure validation was performed using MolProbity within PHENIX (71) using the online server (73). r.m.s.d. analysis and rigid body fitting were performed within PyMOL (Schrödinger). Data collection and refinement statistics are summarized in Table 3. Minimization of the partial BAM model was performed using Chiron (74). All figures were made using PyMOL (Schrödinger) and annotated and finalized in Adobe Photoshop/Illustrator.

Densitometry analysis
Protein abundance was quantified by densitometry using the Image Lab 5.0 software (Bio-Rad) volume tool (rectangle), local background subtraction, and linear regression method as described previously (49,75). Experiments were performed in biological triplicates and are shown in Fig. S3. The relative protein levels are presented as mean values and S.D.

Statistical analysis
GraphPad Prism's built-in t test was utilized to determine statistically significant differences between experimental results. A confidence level of 95% was used for all analyses.   a Values in parentheses are for the highest-resolution shell. b CC1 ⁄ 2 : Pearson's correlation coefficient (CC) between intensity estimates from half data sets. Primary indicator for use for selecting high resolution cutoff for data processing. c Calculated using MolProbity.