Structural insights into outer membrane protein biogenesis in pathogenic Neisseria

Summary

N. gonorrhoeae (Ngo) cause the sexually transmitted infection gonorrhea with ~106 million infections worldwide annually. Ngo infections can result in an increased risk of getting HIV, infertility, and blindness. To combat Ngo infections, we report the cryo-EM structure of the Ngo β-barrel assembly machinery (NgBAM), which is responsible for the biogenesis of β-barrel outer membrane proteins (OMPs). NgBAM was observed in an inward-open state, however, the POTRA domains more closely matches those found in the outward-open state in E. coli BAM. The barrel seam of NgBamA consists of partial pairing of strand β1 with β16; no outward-open state of NgBAM was observed. MD simulations reveal unique overall dynamics and interplay between the POTRA domains of NgBamA and NgBamD. We propose that in Ngo, initial recognition occurs in the inward-open state where the last strand of the OMP partially pairs with β1 of NgBamA and must compete off β16.

eTOC:

BAM is essential for the biogenesis of OMPs in Gram-negative bacteria. Billings et al. report the cryo-EM structure of BAM from Neisseria gonorrhoeae. NgBAM was observed in the inward-open state indicating β16 must be competed off during OMP biogenesis and MD simulations reveal a dynamic interplay between BamA and BamD.

Graphical Abstract

Introduction

Neisseria gonorrhoeae (Ngo) is an obligate human pathogen and the causative agent of the sexually transmitted disease gonorrhea. In the United States alone, ~0.6 million new Ngo infections occur each year and over half of which are from highly drug resistant strains¹ with ~106 million infections expected worldwide annually (World Health Organization; WHO). If left untreated, Ngo infections can also spread to other regions of the body and cause life threatening complications such as infertility, infective endocarditis, and septicemia.^2,3 Ngo has become increasingly resistant to clinical treatments prompting the Centers for Disease Control and Prevention (CDC) to now recommend treatment with only ceftriaxone.¹ Unfortunately, progress towards the discovery and development of novel therapeutic options to combat Ngo infections has been notoriously challenging; both in discovering new antibiotics and particularly with efforts towards a vaccine.^4–6 Therefore, the search for promising new drug targets in Ngo has been given highest priority by the CDC and WHO.⁷

Like other Gram-negative bacteria, Ngo contains both an inner membrane (IM) and outer membrane (OM). The IM contains exclusively α-helical transmembrane proteins, whereas the OM contains almost exclusively β-barrel outer transmembrane proteins (OMPs).⁸ These OMPs are essential for cell survival and virulence, where they are involved in adhesion, import of nutrients from the host, and evasion of the host immune system.^9–13 The biogenesis of OMPs into the OM is mediated by the β-barrel assembly machinery (BAM).^14,15 BAM has long been recognized as an important target for the development of new antibiotics and vaccines, and recent studies describing compounds targeting BAM have further confirmed its promise as a therapeutic target.^16–21

The structure and function of BAM has been widely studied in E. coli, where it consists of five components called BamA, BamB, BamC, BamD, and E. BamA is an OMP itself with five N-terminal polypeptide transport associated (POTRA) domains and a C-terminal 16-stranded β-barrel domain, while BamB, BamC, BamD, and BamE are periplasmic facing lipoproteins anchored to the inner leaflet of the OM.^22–24 BAM has been observed in differing conformational states: the outward-open, where the first and last β-strands of BamA are open in the membrane at ~45° angle and the inward-open state, where the β-strands come together to weakly hydrogen bond.^25–29 These differing conformational states highlight the dynamic nature of the BamA barrel domain, which is essential for proper BAM function. The BAM proteins themselves play important roles in regulating the conformation of BAM.^30–33 The exact mechanism of how BAM folds its substrate OMPs has long been a focus of study. Several models have been proposed based on the past decade of literature.³⁴ In the ‘assisted’ model, BAM allows for OMP folding by only destabilizing the local membrane and energetically allowing OMPs to fold independently.³⁵ In the ‘budding’ model, substrate OMPs use BamA as a template to successively fold strand-by-strand into the membrane. In the ‘swing’ model, BamA is used as a template, but only during the initial interaction with substrate OMPs. Then the partially folded OMP swings into the membrane and folds upon its detachment from BamA.^36,37 More recently, structures of BamA crosslinked to substrate OMPs trapped during folding have been invaluable to our understanding of OMP biogenesis and have provided evidence strongly supporting the budding model. The last strand of substrate OMPs interact with the exposed, first strand of BamA. Then using BamA as a template, they are successively folded eventually bud off into the membrane as a mature protein.^27,34,38

While BamA and BamD are the core and essential components of BAM, the composition and organization of BAM have been shown to vary significantly through evolution.^39,40 Ngo and N. meningitidis distinctly lack an ortholog of BamB (Figure 1A, and Table S1).^41–43 It is unclear how BAM in Ngo and other bacteria that lack one or more of the accessory proteins may adjust their mechanisms to compensate for the lack of critical components. Additionally, a unique putative binding partner, RmpM, was found to co-purify with BAM in N. meningditis.⁴³ The exact role RmpM may have in BAM function, if any, is not well understood. However, RmpM has been shown to stabilize oligomeric OMP complexes in the OMs of Neisseria and may also do so with BAM.⁴⁴

Figure 1. The cryo-EM structure of NgBAM.

(A) The role of NgBAM in the biogenesis of OMPs. (B) Size-exclusion profiles of purified NgBAM and NgBAM_ΔHG2. (C) SDS-PAGE of the purified NgBAM samples. Asterisks indicate degradation products of BamC that were verified using mass spectrometry analysis. (D,E) Orthogonal views of the cryo-EM map and model of NgBAM with NgBamA, C, D, and E colored blue, green, gold, and magenta respectively. (F) The major contacts points between NgBamA and D were interactions with POTRA domains 1,2, and 5. (G) Residue interactions between NgBamA POTRA 5 and NgBamD. (H) Residue interactions between NgBamA POTRAs 1 and 2 and NgBamD. (I) Interactions of the N-terminus of BamE with L2 and L3 of the barrel domain of BamA. (J) Interactions of the extended loop of TPR3 in BamD with L1 of the barrel domain of BamA.

To better understand the structural properties and mechanistic details for how BAM in Ngo mediates OMP biogenesis, we report the cryo-EM structure of BAM from Ngo (NgBAM). NgBAM consists of a central 16-stranded OMP called BamA and three lipoproteins BamC, BamD, and BamE. The β-barrel domain of BamA was observed in an inward-open state, with a significant shift in the rotational angles of the POTRA domains compared to the previously reported structure of NgBamA alone. The barrel seam of NgBamA consists of partial pairing of strand β1 with β16. Molecular dynamic simulations showed unique overall dynamics with a dynamic interplay between the POTRA domains of NgBamA and NgBamD. No outward-open state of NgBamA was observed in either the cryo-EM structure or in the MD simulations. Our studies support a model where NgBAM shares the overall general mechanism for OMP biogenesis with EcBAM; however, the initial recognition step is initiated while NgBamA is in the inward-open state where the last strand of the OMP directly pairs with the unpaired region of β1 of NgBamA and entirely competes off β16. Possible roles for the Ngo accessory proteins are to assist in regulating the conformational plasticity of NgBamA, gate access of nascent OMPs to site of insertion at the lateral seam, and/or gate release of periplasmic domains of mature OMP substrates.

Results

Cryo-EM structure of BAM from Neisseria gonorrhoeae

For structural studies, full length constructs of BamA, BamC, BamD, and BamE were co-expressed using the vectors pET-Duet (BamA/E) and pCDF-Duet (BamC/D), with BamE containing a C-terminal His-tag for purification (Figure S1 and STAR methods). We observed better expression and reproducibility using two Duet vectors compared to a single vector containing all four genes. Following expression, the cells were lysed and the membranes isolated and solubilized with 0.5% DDM. IMAC using the clarified sample was performed using a Ni-NTA column and the peak fractions verified by SDS-PAGE and concentrated, followed by SEC into 1x PBS and 0.01% LMNG. SDS-PAGE was used to confirm the presence of all components (Figure 1B, C). While the complex contained all four proteins at a 1:1:1:1 ratio and appeared stable overall, mass spectrometry analysis confirmed that NgBamC had a propensity for degradation (data not shown). Therefore, we engineered a truncated construct lacking the helix grip 2 (HG2) domain for our studies which had improved yields and purity (NgBAM_ΔHG2). The HG2 domain has only been observed in a few of the many EcBAM structures indicating it does not form stable interaction with other components of the complex³⁴. After truncation of HG2, the SEC profile for the purified complex resulted in a right shifted elution profile (Figure 1B). The HG domains are likely extended in conformation, therefore we speculate this shift is due to the overall hydrodynamic radius of the complex becoming smaller. Cryo-EM grids of purified NgBAM_ΔHG2 were prepared, screened, and data were collected at the National Center for Cryo-EM Access and Training (Figure S2A). Using CryoSPARC, Patch Motion Correction and Patch CTF Estimation were performed, followed by Blob Picker and iterative rounds of 2D Classification until the 2D class averages contained discernable features (Figure S2B). Iterative rounds of Ab Initio Reconstruction, Homogenous Refinement, Non-Uniform Refinement, and Local Refinement produced the best reconstruction of the complex to 4 Å resolution (Figure 1D and Figure S2C and D). The resulting map of the major class was sufficient to unambiguously construct the fully assembled NgBAM using the existing crystal structures of NgBamA, D, and E and a homology model of NgBamC^42,45 (Figures 1E and S3). Unexpectedly, minor classes resulted in lower quality structures of partial complexes with one lacking BamC (NgBamADE, 6.3 Å resolution) and another lacking BamC and E (NgBamAD, 5.0 Å resolution) (Figures S2E and S4 and Table S2). Small conformational changes were observed in NgBamD and NgBamE when compared to their individual crystal structures, with RMSD values of 3.5 Å and 1.5 Å, respectively (Figure S5).

In the fully assembled NgBAM_ΔHG2 structure, NgBamA forms the central scaffold of the complex with NgBamD and E making direct contact with the POTRA domains primarily, with minimal contacts to periplasmic loops of the β-barrel domain (Figure 1F). The TPR repeats 3 and 4 of NgBamD form the most extensive interface with POTRA5 of NgBamA (~1,500 Ų). Notably, D123, D137, Y187, Y188, and R200 of NgBamD are involved in hydrogen bonding and salt bridge interactions with N358, R366, E373, and M372 of POTRA5 of NgBamA (Figure 1G). Included in this interface is the conserved salt bridge formed between E351 of NgBamA and R200 of NgBamD, which has been shown to be important for complex formation and possible activation of BamA.^46,47 The second major contact site involves TPR repeats 1 and 2 of NgBamD which interact along POTRAs 1 and 2 (Figure 1H). Here, R64, T67, R72, and R97 of NgBamD are involved in both hydrogen bonding and salt bridge interactions with R159, Q34, E37, and Q122 of POTRAs 1 and 2. Additionally, the extended loop in TPR repeat 3 of NgBamD extends up and interacts with periplasmic loop 1 of the barrel domain of NgBamA (Figure 1F). NgBamE is observed sandwiched between TPR repeats of 4 and 5 of NgBamD and POTRAs 4 and 5 of NgBamA, with the N-terminus extending up to interact with periplasmic loops 2 and 3 of the barrel domain of NgBamA (Figure 1E). A summary of the inter-protein interactions within the fully assembled NgBAM_ΔHG2 complex structure is provided in Tables S3–S6.

Additional minor 3D subclasses were observed for partial complexes in the cryo-EM dataset, possibly due to destabilization of the complex induced during grid preparation. A comparison of the fully assembled NgBAM_ΔHG2 structure to the minor subclasses for the NgBamADE and NgBamAD structures indicates minimal overall conformational changes (RMSDs ranging from 1.7 – 2.3 Å overall; 1.0 – 1.5 Å for the β-barrel domain only) (Figure 2A, B). The primary conformational shifts observed between the three subcomplexes is along the region were TPR repeats 1 and 2 of NgBamD interact with POTRAs 1 and 2 of NgBamA, where a 7–12 Å shift is observed (Figure 2C). The largest shift was observed with the presence of NgBamE with a minor shift upon the addition of NgBamC. This region also exhibits the lowest local resolution and largest B-factors in the fully assembled NgBAM_ΔHG2 structure indicating increased dynamics and flexibility (Figure S4B, C).

Figure 2. The NgBAM lipoproteins regulate the conformational state of NgBamA.

(A) The NgBamADE and NgBamAD structures are aligned with the full complex NgBamACDE. (B) Minimal differences are observed when aligning the barrel domain individually. (C) POTRA domains individually. (D) The crystal structure of NgBamA alone is aligned to the complex, indicating that with in complex with the NgBAM proteins, the POTRA domains shift to accommodate their interaction. (E) Alignments of the POTRA domains of NgBamA and NgBAM (RMSD= 1.13 Å). (F) Alignments of the barrel domains (RMSD= 2.9 Å). (G) At the lateral gate of NgBAM, the shift in the POTRA domains causes the periplasmic face of the barrel to constrict. Shift of P5 to allow access to the lateral seam and b1 edge. (H) unpaired b1 edge measuring about 15 Å.

A comparison of NgBamA observed in the NgBAM_ΔHG2 cryo-EM structure to the crystal structure of the protein alone (RMSD of 14.6 Å) shows the major differences are along the POTRA domains, which are rotated ~90° away from the central axis of the barrel to accommodate the binding of NgBamD and resulting in a ~40–60 Å sweep of POTRAs 1–4 (Figure 2D). An alignment of just POTRAs 1–5 had an RMSD of 3.5 Å, while an alignment of POTRAs 1–4 has an RMSD of 2.7 Å, indicating a primarily rigid body rotation of the POTRAs between the two structures (Figure 2E). The rotation of the POTRA domains upon NgBamD binding leads to changes in the β-barrel domain (RMSD of 1.4 Å) localized primarily along strands β1-β7, where small angle shifts within the strands result in ~11 Å constriction of the barrel diameter along the periplasmic face (Figure 2F). Strands β1 and β16 have minimal interactions with one another at the barrel seam consisting of only a few hydrogen bonds, while the majority of β1 has an unpaired exposed edge with a length of ~12 Å (Figure 2G).

NgBamC has been shown to be surface exposed,⁴⁸ however, this has not been supported by reported structures of BAM from E. coli.^25,28,29,49 BamC has the lowest identity of the accessory proteins between Ngo and E. coli; no structure has been reported of NgBamC (Table S1). Therefore, a homology model of the N-terminal domain of NgBamC was fit into the cryo-EM map guided by previously reported structures of EcBAM. At lower contour levels, we could observe an additional domain within the map, which was consistent with a homology model of HG1, although it was not well defined (Figures 3A and S6). Unlike in the EcBAM structures where the HG1 domain has been observed interacting with the POTRA 1 of BamA and BamD, HG1 of NgBamC was observed in proximity of the detergent micelle of NgBamA and closer to the membrane interface. Neither domain of NgBamC was observed in either of the minor subclasses of the partial complexes, consistent with our placement of this subunit in the fully assembled complex (Figure S4A).

Figure 3. NgBAM is a hybrid of conformational states seen in EcBAM.

(A) Model of full NgBAM with the HG1 domain of NgBamC included. Global alignments of NgBAM and EcBAM in the inward state (PDB ID: 5D0O) (RMSD= 6.4 Å) (B) and EcBAM in the outward state (PDB ID: 5LJO) (RMSD= 4.7 Å) (C). The conformational changes in these alignments were less prevalent when aligned by the barrel domain (D and E) and the POTRA domains (F and G).

NgBAM is in a hybrid conformational state compared to EcBAM

Based on the conformation of the β-barrel domain of NgBamA, NgBAM_ΔHG2 was observed in a conformation that most closely matches the inward-open state; based on the nomenclature previously used for EcBAM. However, upon further inspection, an overall structural alignment of NgBAM_ΔHG2 to the inward-open state of EcBAM has an RMSD of 6.1 Å, while an overall comparison to the outward-open state of EcBAM has an RMSD of 4.6 Å (Figure 3B, C). And a comparison of the β-barrel domains alone for each of the conformational states resulted in an RMSD of 1.8 Å for the inward-open state (PDB ID: 5D0O) and 3.0 Å for the outward-open state (PDB ID: 5LJO) (Figure 3D, E). Further, the relative position of the POTRAs to the barrel domain align more closely with the outward-open state than the inward-open state globally (Figure 3F, G). The POTRA domains alone do show a significant difference when aligned, with RMSDs of 2.5 Å in the inward-open state and 2.8 Å for the outward-open state. Therefore, NgBAM_ΔHG2 appears to be in a hybrid conformation, sharing properties of both the inward-open state (β-barrel domain of NgBamA) and the outward-open state (the rest of the complex).

NgBAM exhibits a dynamic interplay between BamA and BamD

Molecular dynamics (MD) simulations were carried out for both NgBAM_ΔHG2 and EcBAM (PDB ID 5D0O), as well as on a variant of EcBAM lacking BamB (EcBAMΔB) to enable comparisons since NgBAM lacks a BamB ortholog. Simulations were performed in triplicate for NgBAM_ΔHG2, NgBamA_, EcBAM, and EcBAMΔB, with each at least 6 μs. The RMSD of each domain and their contact area with the membrane were monitored over time to evaluate system equilibration and structural stability across independent trajectories (Figures S7 and S8). After initial relaxation, each trajectory converged to a similar range of RMSD and contact areas, respectively, regardless of starting conformation. The initial conformations of the POTRA domains differed between NgBAM_ΔHG2 and EcBAM with these differences becoming more pronounced over the course of the simulations (Figures 4 and S9). The interaction between the POTRA domains of BamA and BamD were mostly stable in EcBAM and EcBAMΔB, however, this interaction was significantly more dynamic in NgBAM_ΔHG2 and the interaction patterns also varied. In the NgBAM simulations, BamD exhibited a more variable interaction pattern with the POTRA domains compared to EcBAM. While there were instances in which BamD was in contact with both POTRA 1 and 2, similar to the EcBAM simulations, there were also instances in which BamD only interacted with one of the two domains, either POTRA 1 or 2 (Figure 4A, B). This variability in the POTRA-BamD interactions contrasts with the consistent dual-POTRA-domain contact observed in the EcBAM and EcBAMΔB simulations. The contact areas over time between the POTRA domains and BamD for each set of simulations are presented in Figure 4C–E, with the EcBAM and EcBAMΔB simulations having average areas around 470–670 Ų and standard deviations around 100 Ų. The NgBAM_ΔHG2 simulations, however, display significant fluctuations with areas of 325 to 675 Ų in the three simulations, respectively, and standard deviations of ~200 Ų (Figure 4F).

Figure 4. Comparative analysis of the conformational states of BamD and POTRA domains in MD simulations.

Snapshots showing notable conformational states of BamD and POTRA domains observed in MD simulations: (A) for EcBAM, (B) for NgBAM_ΔHG2 with BamD primarily interacting with POTRA1 only (left) and with POTRA2 only (right). BamA and BamD are shown in light blue and orange, respectively (BamB, BamC, and BamE are not shown). (C-E) Moving averages of the contact areas between BamD and POTRA domains from independent 6-μs MD simulations for (C) EcBAM, (D) EcBAMΔB, and (E) NgBAM_ΔHG2. (F) Aggregated statistics of the contact areas observed across the MD simulations, categorized by simulation group; error bars represent the standard deviation across a single simulation trajectory.

The accessory proteins regulate the conformational states of NgBamA

Given the unique dynamics observed in the initial 6 μs of the NgBAM_ΔHG2 simulations, we elected to extend them to 10 μs to further investigate the progression of these dynamics over a longer timescale. The results revealed continuous variability in the contact area between the POTRA domains and BamD, culminating in the complete loss of contact after 6 μs (Figure 5A and B). The barrel domain of NgBamA also adopted distinct conformations compared to EcBamA in EcBAM and EcBAMΔB. Specifically, EcBamA’s lateral gate remained closed with a hydrogen bond between Y432 and I806 (Figure 5C), and the third replica of EcBAMΔB exhibited an open lateral gate with an outward tilt at the top of β1 (Figure 5D). In contrast, by the end of 10 μs in the second NgBAM_ΔHG2 replica, the lateral gate displayed a unique conformation with an inward bend at the end of β16 (Figure 5E), not observed in other groups. These findings highlight the influence of the POTRA domains and accessory proteins on the dynamics of the lateral gate. The lateral gate’s β1 side engages with BamD, while its β16 side is impacted by the POTRA domains, which in turn are influenced by their interaction with other accessory proteins (Figure 5F).

Figure 5. Conformational dynamics of NgBAM_ΔHG2 during extended 10-μs MD simulations.

(A) Moving average of the contact area between BamD and POTRA domains for NgBAM throughout the 10-μs MD simulations. (B) Snapshot capturing a conformational state wherein BamD exhibits no interaction with the POTRA domain. BamA and BamD are shown in light blue and orange, respectively (BamB, BamC, and BamE are not shown). (C-E) Snapshots depicting the lateral gate in different systems: (C) EcBAM with a closed lateral gate, where a single backbone hydrogen bond is formed between Y432 and I806, (D) EcBAMΔB with an open lateral gate, and (E) NgBAM_ΔHG2 with a distorted lateral gate; BamA strands β1 and β16 strands are in purple and violet, respectively. Residues that potentially form hydrogen bonds at the lateral gate are indicated (Y432 and I806 for EcBAM; W432 and L788 for NgBAM). (F) NgBAM_ΔHG2 structure highlighting the contact points along the strand β1 side of the barrel domain of NgBamA and NgBamD and strand β16 side of the barrel domain of NgBamA and the POTRAs to modulate barrel conformations. (G) Summary model for OMP biogenesis by NgBAM, which is in the inward-open conformation during resting state. The exposed β1 edge of the barrel domain of NgBamA serves as a template for the last strand of the nascent OMP, which leads to the folding of the OMP into the outer membrane.

To further analyze the behavior of NgBAM_ΔHG2, we measured the number of hydrogen bonds formed and separation distance at the lateral gate (Figures S10 and S11). No significant difference was seen between NgBAM_ΔHG2, EcBAM, and EcBAMΔB, the lateral gate exhibited a similar number of hydrogen bonds and separation distance among each group. However, when NgBamA alone was additionally subjected to MD simulations, we observed the lateral gate to be widely flexible, as evident by the variable number of hydrogen bonds and fluctuating separation. These results suggest that the NgBAM accessory proteins may regulate the conformation of the NgBamA barrel domain. Furthermore, we measured the rotation angle of POTRA 5 relative to the β-barrel domain in BamA among each of these groups (Figure S9). While the EcBAM and EcBAMΔB simulations showed little variation and flexibility in the rotation angle of POTRA 5 (range of ~20° with SD of ~5°), simulations with NgBAM_ΔHG2 and NgBamA exhibited significant variations with a range of ~−10 – 90° and ~5 – 65°, respectively, with standard deviations of ~30° and ~20°. Further, we analyzed the thickness of the membrane at each strand around the barrel domains (Figure S12). NgBamA alone saw thinning of the membrane around the lateral gate (β1 and β16), which is in agreement with the MD simulations that show flexibility and opening between those strands.⁴⁵ However, when part of NgBAM, similar behavior was observed, indicating that although the NgBAM proteins seem to play a role in regulating the conformation of the POTRA domains, such regulation is not extended into the lateral gate to stabilize it.

Discussion

Ngo is an immediate threat to public health due to the lack of a vaccine and its rapidly growing resistance to clinical antibiotics, underpinning the urgent need for new antibiotics to combat this pathogen. In our study here, we provide structural and mechanistic insights into the Ngo BAM complex, which is an exciting new antibiotic target across all Gram-negative pathogens. We determined the cryo-EM structure of fully assembled NgBAM_ΔHG2, which showed a hybrid conformational state in comparison to EcBAM, with the barrel domain of NgBamA in the inward-open state and the rest of the complex resembling more the outward-open state. Minor subclasses in our cryo-EM dataset yielded additional structures including NgBamADE and NgBamAD. In all the structures with NgBamA, including the previously reported crystal structure, the conformation of the β-barrel domain was in the inward-open state. This indicates that the accessory proteins likely do not regulate this conformational state, although they do impose some local changes in strands β1-β7, which leads to a constriction of the barrel diameter on the periplasmic face.

Conversely however, the association of the accessory proteins does impose significant conformational changes in the POTRA domains of NgBamA. The interaction with NgBamD, which binds along POTRAs 1, 2, and 5 of NgBamA, leads to a ~90° rotational switch of the POTRA domains compared to NgBamA alone, which results in a ~40–60 Å sweep of POTRAs 1–4. The addition of NgBamC and E appears to further stabilize the local conformational plasticity here by imposing additional restraints. In the MD simulations, consistent, transient, and unstable interactions were observed between the POTRA domains of NgBamA and NgBamD, and transient behavior at the lateral gate. This inherent flexibility may aid in the recognition or folding of OMPs into the membrane. Further, this transient behavior is consistent with the lower local resolution and conformational heterogeneity observed in the three cryo-EM subclasses.

Also observed in the MD simulations, the barrel domain of NgBamA exhibited similar behavior to EcBamA and was conformationally stable. Whether or not NgBAM indeed cycles between inward-open and outward-open states remains to be determined. However, NgBamA was not observed to shift to an outward-open state in the cryo-EM analysis or in the MD simulations, despite β1 having minimal interactions with β16 and having an elongated unpaired exposed edge. Taken together, our studies support a model for OMP biogenesis in Ngo where the active state of NgBAM is the inward-open state and OMP biogenesis is initiated by the interaction of the last strand of the nascent OMP with the elongated exposed unpaired edge of β1 of the barrel of NgBamA (Figures 2H and 5G). This differs from EcBAM since a conformational shift is necessary to fully expose β1 of EcBamA to pair with the nascent OMP,^27,38,50,51 however β1 in NgBamA is already largely exposed which negates the need for a conformational switch. Strand pairing between the last strand of the OMP with β1 of NgBamA would fully compete off β16 and the biogenesis of the new OMP would proceed using a strand-templating mechanism in the C-terminal to N-terminal direction. As with EcBAM, the growing OMP would extend away from the NgBamA barrel in a budding manner until it is fully mature.

The role of the accessory proteins may be to assist in exposing β1 of NgBamA given that NgBamD binding induces a drastic shift of the POTRA domains, moving POTRA 5 away from the lateral seam and opening access to the unpaired exposed edge of β1. Similarly, the elevated plasticity in NgBAM compared to EcBAM may also assist in mediating initial recognition of the nascent OMP. A dynamic interplay between NgBamA and NgBamD was discovered in the MD simulations, with one state showing no physical interactions along this region, which may serve as a form of gate. While more studies are needed to fully interrogate these results, possible roles for these observations may be to either (i) allow the nascent polypeptide to access the lateral seam of NgBamA during biogenesis, or (ii) allow the release and passage of periplasmic domains of mature OMPs as they are released by NgBAM; both of which would likely be shared in EcBAM.

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Nicholas Noinaj, (nnoinaj@purdue.edu).

Materials availability

Plasmids and other non-commercially available reagents used in this study are available from the Lead Contact under a material transfer agreement with Purdue University.

Data and code availability

• Cryo-EM maps and models have been deposited into the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) with the following accession codes as noted Table S2: NgBamACDE (EMDB: 45754, PDB ID: 9CMW), NgBamADE (EMDB: 45755, PDB ID: 9CN0), and NgBamAD (EMDB: 47576, PDB ID: 9CN1). They are publicly available as of the date of publication.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.

STAR methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

The genes for cloning originate from Fusobacterium nucleatum ATCC 23726 and cloning was performed as described in the Method Details section. The Escherichia coli strains used for plasmid propagation and expression in this study are listed in the Key Resources Table. The cells were grown as described in the Method Details section.

Key Resources Table.

Bacterial and virus strains
T7Express cells	NEB	CAT# C2566H
Chemicals, peptides, and recombinant proteins
2xYT Media	Research Product International	X15600-5000.0
Ampicillin sodium salt	Fisher Bioreagents	BP1760-25
Streptomycin	Research Product International	3810-74-0
DNAseI	Millipore Sigma	DN25
Phenylmethylsulfonyl fluoride (PMSF)	Gold Biotechnology	P-470-25
Imidazole	Fisher Chemical	O3196-500
n-Dodecyl-β-D-Maltoside (DDM)	Anatrace	D3105-25
Hydrochloric acid (HCl)	Fisher Chemical	A144-212
Lauryl Maltose Neopentyl Glycol (LMNG)	Anatrace	NG310
Isopropyl β-D-1-thiogalactopyranoside (IPTG)	Gold Biotechnology	P-470-25
Deposited data
NgBamACDE structural model	This study	PDB ID: 9CMW
NgBamACD structural model	This study	PDB ID: 9CN0
NgBamAD structural model	This study	PDB ID: 9CN1
NgBamACDE cryo-EM map	This study	EMBD-45754
NgBamACD cryo-EM map	This study	EMBD-45755
NgBamAD cryo- EM map	This study	EMBD-45756
Oligonucleotides
NgC_HG2_trun_SDM_F: GGACGGCAGGCGGAAAACTAATTGGCAAAAAAACC GACC	This study	N/A
NgC_HG2_trun_SDM_R: GGTCGGTTTTTTTGCCAATTAGTTTTCCGCCTGCCG TCC	Ihis study	N/A
Recombinant DNA
pETDuet	Novagen (Millipore Sigma)	71146
pCDFDuet	Novagen (Millipore Sigma)	71340-M
pNgBamAE	This study	N/A
pNgBamCD	This study	N/A
pNgBamCD_ΔHG2	This study	N/A
Software and algorithms
Leginon	Suloway et al52	https://emg.nysbc.org/projects/leginon/wiki/Leginon_Homepage
MotionCor2	Zheng et al53	https://emcore.ucsf.edu/ucsf-softwarech
CryoSPARC	Punjani et al54 Punjani et al55	https://cryosparc.com/
Phenix	Adams, et al60	https://phenix-online.org/
Coot	Emsley et al59	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
PyMOL 3.0	Schrödinger	https://www.pymol.org/
ChimeraX	Pettersen et al56	https://www.cgl.ucsf.edu/chimerax/
Adobe Illustrator	Adobe	https://www.adobe.com/products/illustrator.html
AlphaFold2	Jumper et al58	https://alphafold.ebi.ac.uk/
CHARMM-GUI	Wu et al62	https://www.charmm-gui.org/
NAMD3	Phillips et al71	https://www.ks.uiuc.edu/Research/namd/alpha/3.0alpha/

METHOD DETAILS

Construct design and mutagenesis

The genes for BamA (Omp85), BamC (NlpC/DapX), BamD (ComL), and BamE (SmpA) were amplified from N. gonorrhoeae FA1090 genomic DNA using PCR. The endogenous lipidation signal sequences for NgBamC, BamD, and BamE were replaced with the signal sequence from E. coli BamB (MQLRKLLLPGLLSVTLLSG) to aid the recognition and lipidation of the recombinant lipoproteins. Preceding the gene for NgBamA, a pelB signal sequence was introduced to route expression to the periplasm. NgBamA (residues F23 to F792) and NgBamE (residues C18 to Q125) were subsequently cloned into pETDuet using NcoI and EcoRI and NdeI and XhoI, respectively, to generate pNgBamAE. A 10x histidine affinity tag was introduced on the C-terminus of NgBamE. NgBamC (residues C42 to R398) and NgBamD (residues C17 to H267) were cloned into pCDFDuet using the same enzymes, generating pNgBamCD. A summary of the constructs in this study can be found in the STAR methods. A diagram of constructs used can be found in Figure S1. In order to truncate the HG2 domain of NgBamC, site directed mutagenesis was performed to replace A265 with a stop codon (TAA) to generate pNgBamCD_ΔHG2.

Expression and purification of NgBAM

For expression, 100 ng of pNgBamAE and pNgBamCD plasmid were co-transformed into T7 Express cells (NEB # C2566H) and plated on an LB-agar plate containing 50 μg/mL of streptomycin (strep) and 50 μg/mL of carbenicillin. The plate was incubated overnight at 37°C and a single colony was used to inoculate one 75 mL culture of 2xYT media containing 50 μg/mL of strep and 100 μg/ml of ampicillin (amp). This starter culture was shaken overnight at 37 °C. Afterwards, this culture was used to inoculate 12 liters of 2xYT media also containing 50 μg/mL of strep and 100 μg/mL of amp. Here, 5 mL of starter culture was used for each liter of expression culture. The cells were then grown at 37°C to an OD₆₀₀ between 0.6 and 0.9. IPTG was added to a concentration of 350 μM and the cells were allowed to grow further at 37 °C for 2 to 3 more hours. The cells were harvested from the media by centrifugation of 4000 × g for 15 minutes. The cell pellets were then collected, flash frozen in liquid nitrogen, and stored at −80°C until needed for purification.

For the purification of the complex, all steps were performed either on ice or at 4°C. The cell pellets were thawed and resuspended in 1x PBS pH 7.5 at a ratio of 1 gram of cell pellet per 5 mL of buffer. The resuspended cells were supplemented with phenylmethylsulfonyl fluoride (PMSF) and DNase to final concentrations of 0.2 μM and 10 μg/mL respectively. The cells were lysed with three passages through an Emulsiflex C-3 homogenizer (Avestin) operating at a pressure of 10,000 to 15,000 psi. Debris and unlysed cells were pelleted by centrifugation of 4000 × g and the supernatant was subjected to further centrifugation at 200,000 × g. The membrane pellet was then collected and homogenized in 150 mL of solubilization buffer (1x PBS pH 7.5 and 0.5% DDM, pH 7.5) using a dounce homogenizer. The solution was stirred overnight at 4°C. The solubilized membranes were then clarified by centrifugation at 200,000 × g. Purification was performed using an Atka Pure system (GE Healthcare) with a flow rate of 4 mL/minute. The sample was supplemented with 25 mM imidazole and applied to a 5 mL column packed with HisPur NiNTA resin pre-equilibrated with Buffer A (1x PBS pH 7.5 and 0.05% DDM). The complex was eluted from the column using Buffer B (1x PBS pH 7.5, 0.05% DDM, and 500 mM imidazole) in a 150, 300, and 500 mM imidazole step gradient. The complex was routinely observed to elute at 300 mM imidazole or higher. The peak fractions were pooled and concentrated and injected onto a Superdex200 10/300 GL Increase (Cytiva) size-exclusion column pre-equilibrated with 1xPBS and 0.01% LMNG. Peak fractions were analyzed with SDS-PAGE and fractions containing NgBAM were pooled and concentrated for further use.

Cryo-EM grid preparation and data collection

For structural studies, NgBAM (4 mg/mL) was used immediately after purification. A aliquot of 2.5 μL of the sample was pipetted onto freshly glow-discharged Quantifoil grids (R3.5/1, 200 mesh), incubated for 15 seconds, and blotted using a Vitrobot Mark IV (ThermoFisher) for 0.5 to 3 seconds with a blotting force of 2 at 4°C with 100% humidity. The grids were then plunge-frozen in liquid ethane and transferred to liquid nitrogen for storage. Following screening for suitable ice thickness and particle distribution, data was collected at the National Center for Cryo-EM Access and Training (NCCAT) using a Titan Krios electron microscope (ThermoFisher) operating at 300 kV with a nominal magnification of 81,000x with a K3 direct electron detector (Gatan). Data collection was performed in super-resolution mode using Leginon⁵². The images were recorded at a defocus range of −0.97 to −2.23 μm, with a calibrated physical pixel size of 0.53 Å/pixel. A total of 10,825 movies were recorded; 40 frames per movie over an exposure time of 1.2 seconds, resulting in a total dose of 41.87 e⁻/Ų. Data collection and refinement parameters for the cryo-EM analyses are summarized in Table S2.

Image processing

For the 3D reconstruction, motion correction on all the movie frames was conducted using MotionCor2⁵³ with a binning factor of 2. The motion corrected movies were then imported into cryoSPARC⁵⁴ and Patch CTF Estimation performed. The micrographs were then curated based on quality and CTF resolution-fit estimation, resulting in 9,793. Reference free Blob Picking was then performed, the particles extracted, and used for the initial rounds of 2D Classification until 2D class averages were observed with clearly distinguishable features. These 2D classes were then used as templates for Template Picking, followed by iterative rounds of 2D-classification to generate a final selection of classes with 233,387 total particles. These particles were then used for Ab initio Reconstruction which resulted in three promising classes. Iterative rounds of Heterogeneous Refinement was then performed followed by Non-uniform Refinement⁵⁵ of the most promising classes. This produced the best map of the NgBAM complex to 4.0 Å resolution, representing 95,431 particles. Two other classes were found but has missing regions representing individual components. We identified a second subclass as NgBamADE, which refined to a resolution of 6.3 Å, representing 27,591 particles. We identified a third subclass as NgBamAD, which refined to 5 Å resolution and represented 35,335 particles.

Model building and refinement

The crystal structures of NgBamA, NgBamD, and NgBamE were initially fit into the cryo-EM map using ChimeraX⁵⁶. In order to accommodate the conformational changes in NgBamA, the barrel and POTRA domains were fit separately. Since no structure of NgBamC has been reported, a homology model of NgBamC was generated using trRossetta⁵⁷ and AlphaFold2⁵⁸ and fit into the cryo-EM map using the EcBamCDE structure as a guide. All model building was performed using COOT⁵⁹ and real space refinement performed using PHENIX.⁶⁰ Refinement parameters for the cryo-EM structures are summarized in Table S2. Local resolution calculations were performed in cryoSPARC.⁵⁴ Figures were prepared using ChimeraX,⁵⁶ PyMOL (Schrödinger), and Adobe Illustrator.

Molecular dynamics simulations

The initial configuration for EcBAM was taken from PDB ID 5D0O, with BamB removed to create the EcBAMΔB variant. We note that EcBAM and EcBAMΔB simulations have been described previously.²⁷ For NgBAM, the starting structure was taken from the cryo-EM data described above. These protein structures were integrated into asymmetric outer membranes using the CHARMM-GUI membrane builder.^61,62 The membranes were constructed with a phospholipid inner leaflet and a lipopolysaccharide or lipooligosaccharide (LPS or LOS) outer leaflet, both of which had compositions specific to the respective organisms.⁶³ The compositions of the membranes are summarized in Table S7. All systems were solvated with TIP3P water.⁶⁴ Mg²⁺ ions were added to neutralize the negative charges present on the LPS molecules. Additional K⁺ and Cl⁻ ions were added to maintain the final salt concentration of 150 mM.

Simulations were conducted using NAMD3.⁶⁵ Proteins and lipids were both modeled using the CHARMM36 force field.^66,67 To maintain a stable temperature of 310 K, Langevin dynamics was used with a damping constant γ of 1.0 ps⁻¹. Pressure was kept constant at 1 atm using an anisotropic Langevin piston pressure coupling.⁶⁴ Hydrogen mass repartitioning (HMR) was applied across all models and thus a uniform time step of 4 fs was used for all the simulations.⁶⁸ Each time step included the calculation of bonded interactions and short-range non-bonded interactions within a 12-Å cutoff. For long-range interactions, the Particle-mesh Ewald (PME) method was employed, updated every time step.⁶⁹ VMD was utilized for visualization and analysis of the simulation results.⁷⁰

QUANTIFICATION AND STATISTICAL ANALYSIS

Cryo-EM data collection and refinement statistics are summarized in Table S3.

Highlights:

• The cryo-EM structure of NgBAM reveals an inward-open resting state

• NgBAM is found in a hybrid conformational state compared to EcBAM

• MD simulations reveal dynamic interplay between the POTRA domains of BamA and BamD

• OMPs must compete off β16 of NgBamA to access β1 during biogenesis