We took a multipronged approach to investigate the putative oligomerization state of S. enterica in liponanoparticles (Figure 3, Figure 3—figure supplement 1A). As SMALPs may display high heterogeneity we turned to size-exclusion chromatography (SEC) to remove aggregated material or free SMA polymer that had carried over through initial purification. A small peak was observed corresponding to the void volume of the column, but most of the material eluted in a sharp peak with a retention time of 11.09 min corresponding to a molecular weight of 464 kDa (Figure 3—figure supplement 1B). The predicted molecular weight of the expression construct of WbaP is 61.2 kDa. However, as the SMA and lipid accompanying WbaP in the SMALP may influence separation on SEC relative to globular standards, we sought alternative methods to accurately determine the mass of the WbaP SMALPs in solution. Size exclusion chromatography with multi-angle light scattering (SEC-MALS), mass photometry, and Direct Mass Technology mode (Orbitrap-enabled charge detection mass spectrometry) (Kafader et al., 2019) were utilized to assess the molar mass of WbaP in SMALP. In agreement with the initial SEC experiments, WbaP eluted in a sharp peak on an analytical SEC column in line with MALS (Figure 3A). The polydispersity of this peak was calculated to 1.001, and the molar mass was calculated to 255±9 kDa. Accurate determination of the mass of a species analyzed by SEC-MALS requires a precise assignment of the refractive index increment (dn/dc) of each component of the analyte, however, the refractive index increment (dn/dc) for SMA polymers is under debate in the literature (Hesketh et al., 2020; Thomsen, 2020) Despite the low polydispersity and narrow mass range assigned to this peak, the three-component nature (protein, lipid, and polymer) of WbaP-embedded SMALPs prevents deconvolution of the mass of the protein component. To this end, we turned to Direct Mass Technology mode and mass photometry, techniques that can accurately determine the mass of macromolecules in solution regardless of their composition (Kafader et al., 2019; Olerinyova et al., 2021) When analyzed by mass photometry or Direct Mass Technology mode, WbaP-containing SMALPs were determined to have a molecular weight of ~240 kDa (Figure 3B and C). As the monomer molecular weight of the WbaP construct was calculated to 61.2 kDa, and the SMA along with encapsulated lipids can add a variable amount of mass depending on the solubilized protein, we reasoned that WbaP could exist as a dimer or trimer in SMALP.

To further define the oligomeric state of WbaP in SMALP, we utilized amine-reactive crosslinking probes. By selecting crosslinkers of a defined length, we reasoned that we could covalently trap oligomers and distinguish dimer from trimer using non-reducing sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (Figure 3—figure supplement 2A). A small panel of commonly used crosslinkers was selected to screen against WbaP in SMALP (Figure 3—figure supplement 2B). Of these, the crosslinker dithiobis(succinimidylpropionate) (DSP), with a spacer length of 12.0 Å, readily crosslinked WbaP (Figure 2D, Supplementary file 1A). Crosslinking efficiency was dependent on DSP concentration, with complete crosslinking observed with 5 mM DSP after a 5 min reaction. Crosslinking could be reversed by the addition of a Dithiothreitol (DTT). Fully crosslinked WbaP migrated at ~2 x the molecular weight of the non-crosslinked species, confirming a WbaP dimer in SMALP.

With high confidence in the monodispersity of our sample and the oligomeric state of S. enterica WbaP in SMALP, we continued with CryoEM structure analysis. Prior to this work, only a small screening dataset had been collected on a Talos Arctica instrument (Dodge et al., 2023) Grids were prepared using the highly purified eluent from SEC, movies were collected using a Titan Krios G3i microscope. 2D classes from extracted particles closely matched those observed earlier, and initial ab-initio 3D models displayed features consistent with a liponanoparticle, as well as two distinct protein lobes outside of the membrane. However, various refinement strategies failed to yield even moderate-resolution reconstructions. Reasoning that the auto-generated refinement mask may include SMA and lipid regions, we sought to generate a reasonable model to assist in manual masking during data processing.

We turned to AlphaFold to generate models suitable for docking and refinement (Jumper et al., 2021) Although the prediction of S. enterica WbaP available from the AlphaFold protein structure database has high confidence throughout most of the model (Figure 4A, Figure 4—figure supplement 1A, B), we noted a region of moderate confidence between the DUF and the PGT catalytic domain. In particular, the placement of a predicted β-hairpin motif from Asn251 to Gln268 stood out as anomalous, as this motif was flipped away from the remainder of the protein, approximately 25 Å from the DUF, and ~37 Å from the PGT domain (Figure 4A). As the AlphaFold multimer routine has been reported to successfully predict oligomeric complexes, (Evans et al., 2021; Mirdita et al., 2022) we assessed whether a dimer model of S. enterica WbaP would represent a better starting point for docking and refinement into the CryoEM reconstruction. The dimer was also predicted with high overall confidence (Figure 4B, Figure 4—figure supplement 1A). Unexpectedly, the β-hairpin motif that had poor placement in the monomer model was predicted to mediate domain swapping within the dimer model, (Bennett et al., 1994; Liu and Eisenberg, 2002) with the β-hairpin from chain A continuing a β-sheet structure present in the DUF of chain B (Figure 4B and C). The predicted local distance difference test (pLDDT) score assigned to each amino acid by AlphaFold can be used to judge the quality of a predicted model on a per-residue basis (Jumper et al., 2021; Evans et al., 2021) Amino acids with pLDDT scores >70 are considered to have high confidence (Akdel et al., 2022) For the β-hairpin domain swap region, the average pLDDT score is 78, indicating high confidence in the placement of this motif (Figure 4—figure supplement 1A). Significantly, the prediction of structures of other Lg-PGTs reveals that this β-hairpin domain swap within a homodimer is highly conserved (Figure 4—figure supplement 2).

Based on the earlier success with crosslinking S. enterica WbaP in SMALP, we further investigated the predicted domain swapping using a targeted crosslinking strategy. A set of Cys-pair variants of WbaP were constructed such that thiol-reactive crosslinkers with limited crosslinking length could be used to ‘staple’ the WbaP dimer at the predicted domain swap interface (Figure 5A–C). For these experiments, we employed two different crosslinking strategies. For the first Cys-crosslinking experiment, we utilized Cu phenanthroline to catalyze the direct oxidation of Cys residues to form a disulfide bond between Cys residues within 2.05 Å. For the second strategy, we utilized dibromobimane (bBBr) (Kim and Raines, 1995) to form a covalent crosslink between Cys residues within 4.88 Å. We observed concentration-dependent crosslinking for two discrete sets of crosslinking pairs along the predicted domain swap interface using either Cu phenanthroline or bBBr (Figure 5D and E). No crosslinking was observed for wild-type protein using either crosslinking strategy, demonstrating that there are no native cysteines within crosslinking distance, (Figure 5E) as predicted by the AlphaFold dimer model.

As a final validation of the predicted dimer model, we used crosslinking mass spectrometry (XLMS). Both DSS and tBu-PhoX were chosen for crosslinking, as DSS is commonly used for XLMS, (Chavez and Bruce, 2019; Piersimoni et al., 2022) and tBu-PhoX has membrane-penetrating properties (Figure 6A; Jiang et al., 2022). After crosslinking, WbaP was exchanged into DDM micelles (Hesketh et al., 2020), and subjected to protease digestion. In-solution digestion resulted in fragments covering 93.86% S. enterica WbaP, and 49 unique crosslinking sites were identified (Figure 6B, Figure 6—figure supplement 1). The majority of the DSS crosslinks localized to the solvent-accessible areas of the PGT domain, while several of the tBu-PhoX crosslinks localized to the membrane-adjacent regions at the putative dimer interface (Figure 6C). Inter vs intra-molecular crosslinks were assigned by XMAS (Lagerwaard et al., 2022) based on distance constraints for each crosslinker (Figure 6A). The Cα−Cα distances bridged by either crosslinker should not exceed 35 Å. The observed inter-chain crosslinks agreed with the predicted dimer interface, with Lys273 proximal to Lys146, Lys148, and Lys81. In addition, Lys81 and Lys451 were within crosslinking range (Figure 6C). With targeted crosslinking experiments validating the predicted domain swapping and XLMS validating the dimeric oligomerization state, we proceeded with the AlphaFold dimer model for CryoEM data processing and refinement.

WbaP from S. enterica LT2 (Uniprot: P26406) was purified as previously described (Dodge et al., 2023). Briefly, C43 cells harboring pAM174 (Sjodt et al., 2018) were transformed with a plasmid encoding S. enterica WbaP with an N-terminal SUMO tag – linker – dual-strep tag sequence. Protein was expressed using autoinduction (Studier, 2005) in 0.5 L terrific broth (Tartof, 1987) supplemented with 150 μg/mL kanamycin and 25 μg/mL chloramphenicol. Cells were incubated at 37 °C until the OD₆₀₀ reached ~1.5, and then the temperature was adjusted to 18 ° C and 1 g solid (L)-arabinose was added to each culture. After 18–20 hr expression, cells were pelleted via centrifugation, pellets were transferred to 1-gallon Ziplock freezer bags, manually spread to a uniform thin layer, and frozen at –80 °C.

The region encoding residues 146–272 from S. enterica WbaP was ordered as a synthetic gene, and cloned into pMCSG7 (Eschenfeldt et al., 2009) using Gibson assembly (Gibson et al., 2009). The resulting plasmid was transformed into E. coli Bl21 (DE3), and protein was expressed using autoinduction (Studier, 2005) in 0.5 L cultures supplemented with 100 μg/mL ampicillin at 37 °C. The temperature was adjusted to 18 °C once the OD₆₀₀ reached 1.5, and cultures were incubated for 18 hr. After expression, cells were harvested via centrifugation, transferred to 1-gallon Ziplock freezer bags, spread to a thin layer, and stored at –80 °C.

S. enterica WbaP Cys-pair variants were generated using primers designed in the QuikChange primer tool (Agilent). Successful mutations were confirmed by Sanger sequencing. Forward and reverse primers are shown below.

WbaP was purified in SMALP200 liponanoparticles as previously described (Dodge et al., 2023). Frozen cells were resuspended in buffer A (50 mM HEPES pH 8.0, 300 mM NaCl) at 4 mL per g pellet. Resuspended pellets were supplemented with 2 mM MgCl₂, 0.06 mg/mL lysozyme (RPI), and 0.5 mg/mL DNase I (Millipore Sigma) and incubated on ice for 30 min. Cells were disrupted via sonication (2 × 90 s, 50% amplitude, 1 s on 2 s off), and the cell membrane was isolated via differential centrifugation (Ti-45 rotor, 9000 g 45 min, reserve supernatant, 140,000 g 65 min). Isolated membranes were diluted to 50 mg/mL using buffer A (assessed by UV absorbance at 280 nm), flash frozen in liquid N₂, and stored at –80 °C. Isolated membranes were thawed on ice, and mixed 1:1 (v/v) with a 2% stock solution of SMALP200 (Polyscope) or buffer A and rotated at room temperature for 1 hr. Soluble liponanoparticles were isolated by centrifugation (Ti45 rotor, 160,000 × g 65 min). After centrifugation, pellets were discarded, and the supernatant was flowed over 1 mL Strep-Tactin XT 4flow resin (IBA Biosciences) pre-equilibrated with buffer A. The flowthrough was re-run over the column bed, and the column was washed with 5 mL buffer A. Protein was eluted using 3 mL Buffer A + 50 mM Biotin (buffer B). Protein-containing fractions were identified by UV-vis (Nanodrop) and pooled. Biotin was removed using 3 x tandem 5 mL HiTrap desalting columns (Cytiva) equilibrated with 25 mM HEPES pH 8.0, 150 mM NaCl (buffer C). Protein purity was assessed via SDS-PAGE, and protein concentration was determined via BCA assay (Pierce). Protein was concentrated to 4 mg/mL, and flash frozen in liquid N₂, then stored at 80 °C.

All purification steps were performed on ice. Frozen cell pellets were resuspended in buffer D (50 mM HEPES pH 7.5, 300 mM NaCl, 20 mM imidazole pH 7.5, 5% glycerol) at 4 mL per g cell pellet. Resuspended pellets were brought to 2 mM MgCl₂, 0.06 mg/mL lysozyme (RPI), and 0.5 mg/mL DNase I (Millipore Sigma). Cells were incubated on ice for 30 min. Cells were then disrupted via sonication (2 × 90 s, 50% amplitude, 1 s on 2 s off). Insoluble material was removed via centrifugation (Ti45 rotor, 42,000 RPM 60 min). The supernatant was sterile filtered and loaded onto a 5 mL NiNTA His-Trap column using an Akta FPLC pre-equilibrated with buffer D. Protein was eluted using a linear gradient from 0–100% buffer E (50 mM HEPES pH 7.5, 300 mM NaCl, 400 mM imidazole pH 7.5, 5% glycerol) over five column volumes. Peak fractions were pooled, supplemented with TEV protease at a 1:7 TEV:protein ratio, and dialyzed overnight at 4 °C against 3 L buffer F (25 mM HEPES pH 7.5, 150 mM NaCl, 5% glycerol). Inverse NiNTA purification was used to remove the TEV protease and the cleaved 6x His tag. Flowthrough and low-imidazole wash from inverse NiNTA purification were pooled, concentrated to 5 mL, and injected on a Superdex S200 column pre-equilibrated with buffer F (25 mM HEPES pH 7.5, 150 mM NaCl, 5% glycerol). Peak fractions were analyzed via SDS-PAGE, and fractions containing the S. enterica WbaP DUF were pooled, concentrated to 2 mg/mL, flash frozen in liquid N₂, and stored at –80 °C.

S. enterica WbaP in SMALP was analyzed to assess monodispersity via size exclusion chromatography (SEC) using an Enrich S650 column (Biorad) pre-equilibrated with buffer C. A 500 μL sample was injected, and peak fractions were pooled and concentrated to 10 mg/mL.

A 50 μL aliquot of 0.5 mg/mL S. enterica WbaP in SMALP was injected onto a WTC-030 fused silica column pre-equilibrated with HEPES buffered saline pH 7.5 (HBS). Eluate was flowed through a DAWN MALS detector (Wyatt) and an Optilab differential refractive index detector (Wyatt). Data were analyzed using ASTRA software (Wyatt), and the system was pre-calibrated with a BSA standard.

The S. enterica WbaP in SMALP was diluted to 40 nM in buffer C to a final volume of 50 μL. A CultureWell gasket (Grace Bio-Labs) was attached to a 24 mm × 50 mm glass coverslip (Electron Microscopy Sciences), and 10 μL buffer C was dispensed into a well to acquire focus. A 10 μL aliquot of 40 nM WbaP solution was added to the buffer well to achieve a final concentration of 20 nM WbaP. Following this a 120 s dataset was collected, and the data was analyzed using DiscoverMP (Refyn Ltd). A standard curve was generated using NativeMark unstained protein ladder (Thermo Fisher Scientific).

SMALPs containing S. enterica WbaP were reacted with a panel of lysine-reactive crosslinkers with variable chemistries, solubilities, and spacer lengths. Stock solutions were made of each crosslinker as follows: 50 mM dimethyl adipimidate (DMA) in water, 50 mM dimethyl suberimidate (DMS) in water, 25 mM bis[2-(succinimidyloxycarbonyloxy)ethyl]sulfone (BSOCOES) in dry DMSO, and 10 mM 3,3’-dithiobis(sulfosuccinimidyl propionate) (DTSSP) in water. Crosslinkers were immediately added to 15 µM WbaP in buffer C at a 20–40 fold molar excess. After incubation at room temperature for 30 min, the reaction was quenched by the addition of 1 µL of 1.5 M Tris-HCl, pH 8. Samples were analyzed by Western blot to detect the presence of crosslinked oligomers. Crosslinking efficiency was annotated as high (>30% total protein crosslinked to dimer), moderate (<30% total protein crosslinked to dimer), or none (no protein crosslinked to dimer).

DSP was dissolved in dry DMSO to a stock concentration of 25 mM and immediately added to a final concentration of 0.1–5 mM to 20 µL of WbaP in SMALP. After incubation at room temperature for 30 min, the reaction was quenched by the addition of 1 µL of 1.5 M Tris-HCl, pH 8. The sample was then divided into two aliquots and reducing (+ DTT) or nonreducing loading dye was added to each sample. Crosslinking was analyzed by SDS-PAGE followed by anti-Strep Western blot analysis.

A 2.5 mM stock solution of bBBr in 20% acetonitrile was prepared. bBBr was added to 15 µM WbaP to a final concentration of either 75 or 300 µM. After incubation at room temperature for 20 min, loading dye was added and SDS-PAGE was performed to assess the WbaP oligomerization state.

A 10 mM 1,10-phenanthroline DMSO stock and a 5 mM CuSO₄ aqueous stock were prepared. CuSO₄ and 1,10 phenanthroline were added to 20 µL of 15 µM WbaP to final concentrations of 50 and 100 µM or 200 and 400 µM, respectively. The reactions were allowed to proceed at room temperature for 20 min, followed by the addition of non-reducing loading dye and SDS-PAGE analysis.

Data were collected at the MIT Characterization.nano facility. Optimal freezing conditions were screened using the Vitrobot system (Thermo Fisher Scientific). Grids were clipped and screened on a Talos Arctica G2 (Thermo Fisher Scientific) equipped with a Falcon 3EC camera (Thermo Fisher Scientific). Optimal particles were observed in grid GD5-3, containing 2 mg/mL S. enterica WbaP. Based on these preliminary results, 4481 movies of SMA30-solubilized S. enterica WbaP were collected on a Titan Krios G3i (Thermo Fisher Scientific, Supplementary file 1B) equipped with a K3 camera (Gatan, Inc). Motion correction, CTF estimation, particle extraction, ab-initio reconstruction, and all refinement steps were conducted using Cryosparc V4 Punjani, 2020; Punjani et al., 2017.

Lg-PGTs from T. thermophilus (Uniprot: A0A510HWX9), A. hydrophila (Uniprot: B3FN88), E. coli (Uniprot: P71241), and S. pneumoniae (Uniprot: Q9ZII5) were selected for structure prediction. AlphaFold models were generated as described previously Dodge et al., 2023. Sequences containing two chains of each Lg-PGT were used as input for a local installation of colabfold Mirdita et al., 2022, and predictions were generated using the flags --amber, --templates, --num-recycle 3, --use-gpu-relax activated.

PHENIX version 1.20.1 was used to process the AlphaFold prediction, dock, rebuild, and real-space refine the processed model, and validate the resulting structural model of S. enterica WbaP (Adams et al., 2010; Afonine et al., 2018). UCSF ChimeraX was used to visualize CryoEM volumes (Pettersen et al., 2021). Structural figures were generated using PyMOL (Schrodinger, 2015). The protein topology cartoon was generated using Pro-origami (Stivala et al., 2011). The EMBL PISA server was used to analyze the dimer interface (Krissinel and Henrick, 2007).

S. enterica WbaP in SMALP nanoparticles was prepared in 20 mM HEPES buffer pH 7.0 at a concentration of 0.6 mg/mL. 5 mM disuccinimidyl suberate (DSS) or 2 mM tert-butyl disuccinimidyl phenyl phosphonate (tBu-PhoX)(Thermo Fisher Scientific) in DMSO was added to the solution and incubated for 1 hr at room temperature. Reactions were quenched with 20 mM of Tris-HCl, pH 8.0, for 15 min. N-dodecyl β-D-maltoside (DDM, 10% stock, Thermo Fisher Scientific) was added to the sample to a final concentration of 1% and the mixture was incubated on ice for 30 min. To remove SMA, MgCl₂ (50 mM stock) was added to a final concentration of 4 mM, and the resulting solution was incubated at 4 °C for 1 hr, and then centrifuged at 21,000 × g at 4 °C for 1 hr. The supernatants were transferred to fresh microfuge tubes and diluted 1:1 with 0.1% SDS, 25 mM DTT, and incubated at 50 °C for 1 hr. Chloroacetamide was added to 25 mM and incubated at RT for 30 min in the dark before acetone precipitation overnight at –20 °C. The samples were washed twice with 90% acetone and the pellet was vortexed with 25 mM ammonium bicarbonate until re-solubilized. Enzymatic digestion was carried out with either trypsin in 0.1% Rapigest (Waters) (1:20 ratio) or pepsin (1:50 ratio). The trypsin digestion was stopped after 16 hr with 1% formic acid (FA). Cross-linked peptides were desalted using Pierce peptide desalting spin column (Thermo Fisher Scientific) and dried. The tBu-PhoX cross-linked peptides were enriched as described previously (Jiang et al., 2022).

Samples were separated by reverse phase-HPLC using a Thermo Scientific Vanquish Neo system connected to an EASY-Spray PepMap 75 µm x 25 cm column over a 60 min 3–65% gradient (A: water, 0.1% formic acid; B: 80% acetonitrile, 0.1% formic acid) at 300 nL/min flow rate. The crosslinked samples were analyzed on the Orbitrap Eclipse Tribrid mass spectrometer with Instrument Control Software version 4.0. 0. Cross-linked samples were analyzed using an HCD-MS2 acquisition strategy with 30% normalized collision energy (NCE). MS1 and MS2 scans were acquired in the orbitrap with a respective mass resolution of 60,000 and 30,000. MS1 scan range was set to m/z 375–1,400 at 100% AGC target, 118 ms maximum injection time, and 60 s dynamic exclusion. MS2 scans were set to an AGC target of 200%, 70ms injection time, isolation window of 1.2 m/z. Only cross-linked precursors at charged states +3 to+8 were subjected to MS2.

Single ion measurement using Direct Mass Technology mode was performed on Thermo Scientific Q Exactive UHMR. Prior to Direct Mass Technology mode MS analysis, sample was buffer exchanged into 100 mM ammonium acetate using Zeba spin column (7 kDa MWCO, Thermo Fisher Scientific). Sample was loaded into Au/Pd-coated borosilicate emitters (Thermo Scientific, ES388) for nano-ESI. For Instrument parameters, ion transfer target m/z and detector optimization were set to ‘high m/z.’ In-source trapping was set at −100 V. Trapping gas pressure readout was ~2 × 10⁻¹¹ mbar. Data were acquired at 200 K @m/z400 resolution setting. Data analysis was done by STORIBoard software (Proteinaceous).