Characterization of the ABC methionine transporter from Neisseria meningitidis reveals that MetQ is a lipoprotein

NmMetQ is a substrate binding protein (SBP) from Neisseria meningitidis that has been identified as a surface-exposed candidate antigen for meningococcal vaccines. However, this location for NmMetQ challenges the prevailing view that SBPs in Gram-negative bacteria are localized to the periplasmic space to promote interaction with their cognate ABC transporter embedded in the bacterial inner membrane. To address the roles of NmMetQ, we characterized NmMetQ with and without its cognate ABC transporter (NmMetNI). Here, we show that NmMetQ is a lipoprotein (lipo-NmMetQ) that binds multiple methionine analogs and stimulates the ATPase activity of NmMetNI. Using single-particle electron cryo-microscopy, we determined the structures of NmMetNI in the absence and presence of lipo-NmMetQ. Based on our data, we propose that NmMetQ tethers to membranes via a lipid anchor and has dual function/topology, playing a role in NmMetNI-mediated transport at the inner-membrane in addition to moonlighting functions on the bacterial surface.


24
The substrate binding protein NmMetQ from the human pathogen Neisseria meningitidis has been 25 identified as a surface-exposed candidate antigen for the meningococcal vaccine (Pizza et al., 2000). 26 Subsequently, NmMetQ was shown to interact with human brain microvascular endothelial cells 27 (Kánová et al., 2018), potentially acting as an adhesin. However, the surface-topology of NmMetQ 28 challenges the prevailing view that substrate binding proteins (SBPs) reside in the periplasm of Below the schematic are the theoretical masses for the lipo-NmMetQ proteins (in italics) assuming triacylation occurs via the canonical lipoprotein maturation pathway, due to the sequential action of three enzymes (Lgt, LspA and Lnt). The numbers in the brackets correspond to the total number of carbons and double bonds, respectively, present in the fatty acyl chains of the lipid. (Bottom) Schematic illustrating various NmMetQC20A proteins with example theoretical average masses, shown in italics, assuming cleavage occurs between A19 and A20, possibly by signal peptidase I (SPase I). N-terminal signal peptides are represented by a green rectangle. B. Characterization of lipo-NmMetQ. Size-exclusion chromatogram and mass spectra of peak 1. The molecular masses of the major species correspond within 1 Da to the predicted mass for two triacylated NmMetQ species, one with acyl chain composition [16:0, 16:0, 16:0] (31,661 Da) and the other with [16:0, 16:0, 18:1] (31,685 Da). C. Characterization of NmMetQC20A. Size-exclusion chromatogram and mass spectra of the major species from peak 2 and peak 3. The molecular masses of the major species of peak 2 and 3 correspond to the pre-protein NmMetQ (32,802 Da) and secreted NmMetQ (30,839 Da), respectively. These measured masses are within 3 Da of the predicted masses for each species. Assigned NmMetQ species are depicted in cartoon form on the chromatograms.   In this work, we characterized NmMetNI and NmMetQ using multiple biophysical methods. Us-97 ing mass spectrometry and site-directed mutagenesis, we demonstrate that full-length NmMetQ, 98 recombinantly-expressed in E. coli, is a lipoprotein (lipo-NmMetQ). Functional assays show that 99 both lipo-NmMetQ and L-methionine are required for maximal stimulation of NmMetNI ATPase 100 activity. NmMetNI can also be stimulated, although to a lower extent, by L-methionine and pre-101 protein NmMetQ (a variant with an unprocessed N-terminal signal peptide), and lipo-NmMetQ 102 and select methionine analogs. We also determined the structures of NmMetNI in the absence 103 and presence of lipo-NmMetQ to 3.3 Å and 6.4 Å resolution, respectively, using single-particle elec-104 tron cryo-microscopy (cryoEM). Using a bioinformatics approach, we also identified MetQ proteins 105 from other Gram-negative bacteria that are predicted to be modified with lipids. This analysis sug-106 gests that the lipid modification of MetQ proteins are not restricted to N. meningitidis and E. coli. 107 Based on our data, we propose that lipo-NmMetQ, and more generally lipo-MetQ proteins in 108 other Gram negative bacteria, have a dual function/dual topology: ABC transporter-dependent 109 roles at the IM and a moonlighting ABC transporter-independent role (or roles) at the OM. Our   146 To confirm that lipid attachment site occurs at the N-terminal Cys 20 on NmMetQ, we gener-147 ated a Cys-to-Ala NmMetQ mutant (NmMetQC20A). We hypothesized that this mutation would 148 prevent lipid attachment and lead to the accumulation of pre-protein NmMetQ, containing an un-149 processed N-terminal signal sequence and the C20A mutation. The NmMetQC20A protein was 150 expressed and purified in DDM as previously described. The SEC elution profile reveals two major 151 peaks with distinct elution volumes, 78 ml and 100 mL for peak 1 and 2, respectively Figure 1C. 152 For peak 1, analysis of the fraction containing the highest peak revealed a deconvoluted mass of respectively. The production of the secreted NmMetQ was surprising since we only expected the  that the R ℎ values and molecular weight estimates were larger for lipo-NmMetQ (R ℎ = 7.9 ± 0.17 nm, Mw-R = 430 ± 22 kDa) and pre-protein NmMetQ (R ℎ 7.7 ± 0.055 nm, Mw-R = 400 ± 6.7 kDa), than for 174 secreted NmMetQ (R ℎ 3.0 ± 0.013 nm, Mw-R = 43 ± 0.33 kDa) (Figure 1-Figure Supplement 1). Based 175 on both the size-exclusion chromatograms and DLS data, we propose that both lipo-NmMetQ and 176 pre-protein NmMetQ aggregate to form micelles-like complexes.   The same protocol was performed with pre-protein NmMetQ, which contains an N-terminal signal 196 sequence, but without the lipid modification. Addition of pre-protein NmMetQ also led to stimula-197 tion of ATPase activity, although to a lesser extent than observed for lipo-NmMetQ (orange trace).

198
Addition of secreted NmMetQ, however, had little effect on the ATPase activity (cyan), however.

199
Together, these data establish that the lipid moiety of lipo-NmMetQ is required for maximal Nm-      To optimize the FAXS experiment, we considered several factors. As shown in Figure 1  To determine whether methionine analogs could serve as potential substrates for the lipo-302 NmMetQ NmMetNI system, we then measured NmMetNI ATPase activity in presence of lipo-NmMetQ 303 and several methionine analogs. For these assays, we chose several methionine analogs identi-304 fied by FAXS to bind NLM-NmMetQ with an affinity similar or higher than D-methionine, a known 305 substrate for E. coli NmMetNI. Since substrate stimulated ATPase activity is a hallmark of ABC 306 transporters (Bishop et al., 1989; Mimmack et al., 1989), we expected methionine analogs that 307 are substrates for this system would stimulate NmMetNI ATPase activity. Figure 2 shows the re-308 sults for the methionine analog stimulation of NmMetNI ATPase activity. As a negative control, we 309 tested L-cysteine, where, as expected, no substrate-stimulated ATPase stimulation was detected.

310
Our data shows that the following methionine analogs led to substrate-stimulated ATPase activity:  We also determined the single-particle cryoEM structure of DDM solubilized NmMetNI in com-   362 We used a bioinformatics approach to determine if other Gram-negative bacteria could have lipid-  NmMetQ has been previously identified as an OM surface-exposed candidate meningococcal vac-374 cine antigen (Pizza et al., 2000), possibly playing a role in bacterial adhesion to human brain en-375 dothelial cells (Kánová et al., 2018). However, the presence of NmMetQ at the OM challenges the 376 prevailing view that SBPs reside in the periplasm, freely diffusing between the IM and OM (Thomas 377 and Tampé, 2020). To better understand whether NmMetQ has lost its ABC transporter-dependent 378 function at the IM and how NmMetQ remains at the surface of the bacterium, we used multiple 379 biophysical techniques to characterize the structure and function of NmMetQ and NmMetNI. Here, 380 we show that NmMetQ is a lipoprotein that binds and stimulates NmMetNI.

381
Based on our data, we propose a model for NmMetQ localization that reconciles previous 382 studies identifying NmMetQ as a surface-exposed candidate antigen and our study characterizing   Our identification of NmMetQ as a lipoprotein is predicated on our ability to express and purify 391 lipo-NmMetQ and its processing variants. We recognize that a key assumption in our study is that EcMetNI-EcMetQ (Nguyen et al., 2018). Together, our data suggests that lipo-NmMetQ plays a role 407 in NmMetNI-mediated nutrient acquisition.  While previous studies have shown that many SBPs of Gram-negative are soluble (Heppel, 1969), 426 our findings suggest that at least some SBPs may be modified with lipids. Since lipid modifications 427 may allow for SBPs to have a surface-topology in Gram-negative bacteria, we believe that future 428 efforts should be made to experimentally determine which SBPs have lipid modifications, dual 429 topology, and ABC transporter-independent functions. Studies aimed at determining the rules that 430 govern protein surface-exposure will not only increase our understanding of bacterial physiology, 431 but will also help in the rational design of vaccines based on surface-exposed protein antigens. MetI contained no additional residues. A similar strategy was used to produce other ABC trans-440 porters. (Locher et al., 2002; Pinkett et al., 2007). To produce lipo-NmMetQ, the DNA sequence 441 encoding the NmMetQ with the native signal sequence and a C-terminal decahistidine tag was 442 added to a single modified pET vector. This construct served as a template to generate the C20A 443 mutant, which was created using PCR site-directed mutagenesis. NLM-NmMetQ was created as 444 previously described (Nguyen et al., 2019). 445 All proteins were expressed in E. coli BL21 (DE3) gold (Agilent Technologies, Santa Clara, CA)   were used to estimate CTF parameters using Patch CTF job in cryoSPARC. Micrographs containing 477 either ice or poor CTF fit resolution estimations were discarded. A subset of images was randomly 478 selected and used for reference-free particle picking using Blob picker. Particles were subjected 479 to multiple rounds of 2D classification, and two classes (top and side) were used as templates for 480 particle picking on the full set of images. The subsequent processing steps were different for the 481 two datasets.

482
For the dataset acquired for NmMetNI in the inward-facing conformation, initial particle stacks 483 were extracted, downsampled four times, and then subjected to 2D classification. Classes that 484 were interpreted as junk were discarded. The selected particles were then used to generate ab 485 initio volumes. Two volumes, interpreted as NmMetNI and a junk/noise class, were selected for 486 heterogeneous refinement. Particles assigned to the NmMetNI class were processed further by 487 repeating the same strategy using particles downsampled twice, and then again with no downsam-488 pled particles. The final resulting particle stack was then non-uniformly refined (Figure 4-Figure   489 Supplement 3). 490 For the dataset acquired for the lipo-NmMetQ:NmMetNI complex in the outward-facing con-491 formation, initial particle stacks were extracted, downsampled ten times and subjected to 2D clas-492 sification. Classes that were interpreted as junk were discarded. 2D classification was then re-493 peated with particles downsampled by four, and then again with no downsampled particles. The

494
Particles assigned to the lipo-NmMetQ:NmMetNI complex class were subjected to another round of ab initio, followed by heterogeneous, refinement. The final resulting particle stack was then 498 non-uniformly refined (Figure 4-Figure Supplement 4). 499 To build the atomic model of NmMetNI in the inward-facing structure, the structure of EcMetNI 500 (PDB: 3TUJ) lacking the C2 domain was used as template for model building. The model was built 501 by rigid-body docking, homology modeling, and manually building into the 3.3 Å resolution cryoEM 502 density in Coot v0.9.1 (Emsley et al., 2010) and refined using ISOLDE (Croll, 2018). The model of 503 the NmMetNI:lipo-NmMetQ complex in the outward-facing conformation was built by rigid-body 504 refinement of NmMetNI in the inward-facing conformation (traced from the 3.3 Å resolution re-505 construction) and of the previously determined soluble NmMetQ structure in the substrate free 506 conformation (PDB:6CVA) were used as template for model building. The model was built by rigid-507 body docking in Coot, followed by refinement in ISOLDE using adaptive distance restraints.

508
Intersubunit distances between ATP-binding domains were defined by the positions of C of 509 glycine residues of the P loop and signature motifs like previously described (Kadaba et al., 2008). 510 Specifically, Gly44/Gly144 and Gly43/Gly143 for NmMetNI and EcMetNI (3TUJ), respectively and  Protein sequences were obtained through the UniProtKB database using the following search 551 terms: Proteobacteria (taxonomy ID 1224), InterPro family IPR004872 (which NmMetQ UniProt 552 ID Q7DD63 is a member) and identity 90%, which groups sequences with > 90 % identity and 80 553 % sequence length. SignalP 5.0 was used separately to analyze the N-terminal protein sequences 554 and predict the location of the signal sequence cleavage sites. Sequence alignment data was gen-555 erated by the EFI Enzyme Similarity Tool (https://efi.igb.illinois.edu/efi-est/) using Option C with 556 FASTA header reading (Gerlt et al., 2015). A SSN network was then created using an alignment 557 score corresponding to approximately 60% sequence identity and filtering for sequences between   Representative DLS intensity distribution plots of lipo-NmmetQ (0.7 mg/ml), pre-protein (2.3 mg/ml) and secreted Nm-MetQ (2.7 mg/ml) (top panel). The hydrodynamic radius (R ℎ ), the polydispersity (Pd %), molecular weight estimate based on the hydrodynamic radius of a folded globular protein (Mw-R) are listed below each plot. The mean and SEM of each measurement were calculated from triplicate measurements. Proposed models of NmMetQ protein quaternary arrangements (bottom panel)     Representative cryoEM micrograph of (scale bar is 20 nm) and select 2D class averages. B. Workflow of single-particle image processing. C. Angular distribution calculated for particle projections. Heatmap shows number of particles for each viewing angle (top) and gold-standard Fourier shell correlation (FSC) curves for masked and unmasked maps generated by cryoSPARC non-uniform refinement (bottom).