Structure of the essential inner membrane lipopolysaccharide-PbgA complex

The following content is a look into the results of the 2020 paper published in Nature, “Structure of the essential inner membrane lipopolysaccharide-PbgA complex” by Clairfeuille et al. and its future implications. The paper discusses the role of PbgA in the LPS biosynthesis regulation and has key implications for novel antibiotic research in the future.

In order to demonstrate and deduce the essentiality of the PbgA protein within the context of the outer membrane, a number of bacterial strains and plasmids were constructed. For this experiment, a uropathogenic Escherichia coli (UPEC) strain was developed with a pbgA deletion (ΔpbgA). This was done using λ Red recombination. The vector of choice was the pKD4 plasmid, specifically the kanamycin or gentamicin cassettes, which were amplified with primers. The results were then transformed into the UPEC strain of interest containing pSIM18. They were then recovered for 4 hours at 37 °C and selected on a medium containing kanamycin, chloramphenicol, or gentamicin. The UPEC-ΔpbgA strain resulted in a single clone and the ΔpbgA mutation was confirmed using PCR and sequencing.

After recovering the colony forming units (CFU) from the thigh muscle of neutropenic CD1 mice, the UPEC-ΔpbgA strain shows a significantly lower number of CFUs recovered ^[1]. This indicates that the lack of PbgA protein impacted the growth of the bacteria, indicating the essentiality of pbgA.

In order to test the differences in growth of the ΔpbgA strain in comparison to the wildtype UPEC strain, the sensitivity to 50% human serum was tested. The wild-type UPEC bacteria exhibited about 10¹⁰ CFU/mL in 50% human serum ^[1]. The strain of the same bacteria with the pbgA deletion, however, forms only about 10⁶ CFU/mL in 50% human serum ^[1]. When this UPEC-ΔpbgA strain is supplemented with additional pbgA, the CFU/mL in 50% human serum increases to about 10^{9 [1]}. To generate the pbgA that was used for this supplementation, pbgA was amplified from UPEC CFT073 and cloned into the pBAD vector using Gibson assembly.

In order to conduct a rifampicin sensitivity assay of the UPEC and UPEC-ΔpbgA strains, the strains were plated onto LB agar plates containing 0.2% glucose and 50 μg mL^-1 carbenicillin and then incubated overnight at 37°C. Three isolated colonies were then re-streaked onto similar LB agar plates with the same glucose and carbenicillin compositions. Then, a single isolated colony for each plate was heavy-streaked onto LB agar containing just the carbenicillin. Bacteria were then scraped, diluted to OD600 0.025, and grown to mid-log phase for about 2.5 hours or approximately 4 generations at a temperature of 37°C. Then, they were back-diluted in fresh LB media to OD600 0.025 and grown to the mid-log phase again. They were then added to the rifampicin assay plates which were made by serially diluting rifampicin stock in LB medium in clear-round bottom 96-well plates. The bacteria were added to each well to a final OD600 0.01, and the plates were intubated at 37°C and read after 6 hours. The dose-response curve was fit using PRISM software. The curve demonstrated greater rifampicin sensitivity in the UPEC-ΔpbgA strain as compared to the UPEC strain, whereas the UPEC-ΔpbgA strain supplemented with pbgA demonstrated greater sensitivity than the UPEC strain, but significantly lesser sensitivity than UPEC-ΔpbgA ^[1]. This established that the strain devoid of pbgA was sensitized to large antibiotics that normally cannot penetrate the outer membrane of UPEC ^[1].

A ΔpbgA strain was then constructed using E.coli K-12 rather than UPEC using similar methods. This ΔpbgA::pBADpbgA was created by first cloning pbgA into the pBAD28 plasmid, and then inserting pBADpbgA at the attB site in the BW35113 E.coli strain, after which the native copy of pbgA was deleted. ΔpbgA::pBADpbgA demonstrated an absence of the suppressor mutation ^[1]. This absence of pbgA resulted in inhibition of growth, as tested by K-12 ΔpbgA::pBADpbgA cultures diluted in a fresh medium. The ΔpbgA::pBADpbgA shows a significantly lower curve that flattens quickly plotted on a time vs. OD600 graph in comparison to the ΔpbgA::pBADpbgA strain that is coupled with an inducer, which is 0.02% arabinose ^[1].

Images taken of E.coli K-12 ΔpbgA::pBADpbgA cells grown without arabinose at about 4 hours indicated an increased cell diameter ^[1]. This suggests a loss of shape in the pbgA deficient bacteria, as a result of a compromise in the outer membrane integrity, which was also observed by membrane bursting.

In order to examine any significant findings in lipid homeostasis levels, analyses of lipid A extracts from outer membrane vesicles were conducted. Kdo₂-lipid A was isolated using a hydrolysis buffer and a boiling water bath that cleaved the O-antigen from Kdo₂-lipid A. Lipid A was then extracted by the addition of 4.5 mL of chloroform and 4.5 mL of methanol. The solvents were then vortexed and separated by centrifugation. After a second round of extraction using 4.5 mL of chloroform, the bottom layers were combined, dried, and dissolved into methanol-chloroform for MALDI-TOF analysis using the 4800 plus MALDI-TOF/TOF Analyzer. Samples were prepared by plating 0.5 μL of the matrix, which was a saturated solution of 6-azra-3-thiothymine in 100% methanol, followed by 0.5 μL of the sample solution. The spectra were acquired in the negative ion reflector mode. The MALDI-TOF analysis showed increased hepta-acylated lipid A species, which indicated disturbed lipid homeostasis ^[1]. Moreover, there was a significant decrease in the total lipid A:phospholipid ratio in the ΔpbgA as compared to the wild-type strain ^[1]. A disturbed ratio greatly compromises outer membrane integrity, indicating the essentiality of pbgA ^[1].

In order to test the effects on a strain devoid of cardiolipin, a triple mutant ΔclsABC strain was constructed by sequentially introducing each cls deletion. These were introduced into E.coli by P1vir transduction and the deletions were confirmed using PCR. This strain was not sensitized to rifampicin ^[1].

Membrane phospholipids were extracted out of the outer membrane vesicle using a modified Bligh-Dyer protocol. The process was repeated, and extracts were combined as well as dried. The dried residues were then reconstituted in 50 μL of 50:50 dichloromethane:methanol along with 10 mM ammonium acetate. Following this, 30 μL of the sample was injected into a MetaSil AQ C18 column on an HPLC system. The flow rate was 0.3 mL/min, and the gradient was held at 40% mobile phase A (methanol containing 10 mM ammonium acetate) for the first 3 minutes. Mobile phase B (dichloromethane and 10 mM ammonium acetate) was increased to 85% over 9 minutes and was then increased to 95% in 30 seconds and was maintained at 95% for 4 minutes before being returned to initial conditions. The HPLC system was coupled with a 6500+ QTRAP mass spectrometer. The bacterial membrane lipids phosphatidylethanolamine, phosphatidylcholine, and cardiolipin were detected via characteristic head group ions that were present upon fragmentation. Phosphatidylethanolamine and phosphatidylcholine were scanned for neutral loss in positive polarity losses of 141 Da (PE) and 184 Da (PC), and cardiolipin was detected through precursor ion scan in negative mode.

When purified in mild detergent, PbgA was found to be stabilized by anionic phospholipids, including phospholipids that are not naturally found abundantly in E.coli such as phosphatidic acid (PA) and phosphatidylserine (PS) ^[1]. This can be seen in figure 1, which showcases the thermostability of purified PbgA. E.coli PbgA can be seen to be highly stabilized by exogenous lipids, those that are not naturally abundantly found in E.coli.

Figure 1: Thermostability of purified //E.coli// PbgA was analyzed using differential scanning calorimetry with or without 0.1 mg mL^-1 lipid supplementation.

Next Clairfeuille et al. crystallized PbgA in lipidic cubic phases. Crystallization was accomplished by setting up crystal screens in LCP using 40 mg/mL PbgA as well as a monoolein. Protein-lipid mixes were prepared at room temperature, and crystals were grown in 50 nL drops, surrounded by an 800 nL reservoir solution.

Rounds of optimization in MemMeso HT screens produced the best diffracting PbgA crystals that were then obtained in a buffer constituting 0.1 M Tris, 0.2 M ammonium sulfate, and 40% PEG200 at 4°C, and grew to their maximum size in roughly 20 days. The crystals were flashed frozen and 180° x-ray diffraction data was collected from a single crystal. PbgA structure was determined using molecular replacement via PHASER. Rigid body refinement of the periplasmic domain template allowed for clear visibility of electron density for the transmembrane domain (TMD). The model was completed and rebuilt via iterative refinement and omitted maps using COOT and PHENIX. All structural figures were generated using the software PyMOL, conservation analysis was performed using Consurf, and structural homologs were explored using the Dali server. Density maps were calculated to 2.0 Å with Fo − Fc maps calculated before the inclusion of LPS into the refined model.

Crystallization allowed for high-resolution structure determination and revealed several extra densities around the TMD ^[1]. This can be seen in data figures 2 and 3. Figure 2 showcases two structures. On the left of figure 2, from PbgA crystallized in space group C2 and using data to 2.0 Å, a Fo − Fc map calculated shows bilobal extra electron density along the periplasmic membrane leaflet, before LPS inclusion. On the right, a Fo − Fc map was calculated from PbgA crystallized in space group P3 before the inclusion of LPS into the model.

Figure 2: Biophysical and structural characterization of PbgA. PbgA crystallized in space group C2 on the left using data to 2.0 Å and a //Fo − Fc// map calculated shows bilobal electron density along the periplasmic membrane leaflet. Right, PbgA crystallized in space group P3, using data to 4.6 Å, and //Fo − Fc// map was also calculated.

Figure 3 reveals representative non-protein densities found surrounding the TMD of PbgA that were assigned as phosphatidylethanolamine or monoolein lipids.

Figure 3: Biophysical and structural characterization of PbgA. Representative non-protein densities that were observed to be surrounding the TMD of PbgA were assigned as putative phosphatidylethanolamine or monoolein lipids. Also included is a //Fo − Fc// map that was calculated prior to the inclusion of phosphatidylethanolamine or monoolein into the model.

PbgA has 5 N-terminal TMD helices at which the C-terminal periplasmic domain sits, shown in figure 4a ^[1]. This figure shows the PbgA crystal structure and electrostatic representation. The interfacial domain (IFD) of PbgA is a compacted 3-helix bundle that connects the TMD to the PD. Moreover, there are considerable amounts of interdomain contacts which suggests that TMD, PD, and IFD are tightly fused ^[1]. Figure 5 shows the interdomain surface area contacts within PbgA.

Figure 4: PbgA’s structural features. a) PbgA crystal structure in cartoon (left) and electrostatic (right) representation. TMD (blue), IFD (pink), PD (green), and LPS shown in green sticks. b) TMD-based alignment with the PbgA–IFD in pink.

Figure 5: Biophysical and structural characterization of PbgA. Illustration of interdomain contacts within PbgA.

Next, an all-atom model of PbgA in a lipid environment was generated via the high-resolution crystal structure using maestro. The protein preparation function was used to add missing residues, side chains, and protons, predict residue protonation states, and optimize side-chain conformations. Two simulations were constructed under the System Builder. An LPS-PbgA simulation containing LPS, proteins, lipids, water, and ions and a PbgA-only simulation that excluded LPS. Each simulation was placed in a POPE lipid bilayer and the system was neutralized. The resulting systems were equilibrated using the relax_membrane.py and Desmond multisim programs. Following equilibration, two production NPT simulations, LPS and PbgA and PbgA alone, were run for 500 ns using DESMOND and 2 additional simulations were run with LPS swapped. Additional simulations were run to assess whether the simulations had reached equilibrium.

Using the crystal form, molecular dynamic studies and comparison to recent structure, revealed no substantial conformation changes (figure 6) ^[1]. Figure 6 showcases the structural superposition of PbgA crystal structures that were determined in this study, as well as chain A and chain B from PDB 6V8Q. The overall root mean square deviation for the main chain atoms between the most divergent structures is less than 0.8 Å ^[1].

Figure 6: PbgA structural alignments and molecular dynamics simulations. Structural superposition of PbgA crystal structures determined in the study (space group C2 and P3) and both chains A and B from PDB 6V8Q.

These findings indicate that the PD remains anchored onto the TMD and only protrudes 60 Å above the inner membrane ^[1]. This can be seen in figure 4.

These findings oppose a previously proposed model, the cardiolipin-transporter model ^[1]. This model suggests that the periplasmic domain shuttles across the periplasm which is around 200 Å.

Furthermore, the IFD is not just a simple linker. The cardiolipin binding site hypothesized to be within the PD is distant from the inner membrane, and most probably cannot grant phospholipid access ^[1]. On the right of figure 7 are views of the previously proposed cardiolipin binding site. The residues that are predicted to be involved in cardiolipin binding are represented as orange sticks and can be seen to form an integral part of the hydrophobic core.

Figure 7: PbgA structural alignments and molecular dynamics simulations. The previously proposed cardiolipin binding site is shown on the right as orange sticks.

Also, the PD contains no recognizable sequence or structural homology to previously established lipid-binding modules ^[1].

PbgA is structurally related to a superfamily of enzymes that works by modifying the cell envelopes of both gram-negative and gram-positive bacteria. The periplasmic domain is similar to the Mg²⁺ dependent enzyme, LtaS, which synthesizes a surface polymer in S.aureus, which lacks an outer membrane. LtaS is also a membrane protein that has five TMDs and a structure that is similar to PbgA. Figure 8 presents the structure-based alignment of the hydrolase superfamily domains from PbgA, S.aureus LtaS, and E.coli PE transferase MCR-1⁷¹.

Figure 8: PbgA structural alignments and molecular dynamics simulations. Structure-based alignment of the hydrolase superfamily domains from PbgA (PD, green), //S.aureus// LtaS (ECD, blue), and //E.coli// phosphoethanolamine transferase MCR-171 (PD, purple).

Furthermore, full-length PbgA is most similar to the Zn²⁺ dependent, inner membrane-anchored enzyme EptA, which imparts resistance to polymyxins by transferring a PE moiety onto lipid A. Although it can be seen that the PDs and TMDs superimpose well, the compacted α-helical IFD of PbgA exists as an extended linker in EptA and overall structures are very divergent. Figure 9 shows the structure-based alignment of EptA-isolated PDs on the left and TMDs on the right. PbgA does not conserve the side chains required to coordinate Zn²⁺ and mutations within its vestigial active site don't affect outer membrane activity ^[1]. So, the PD appears to be a pseudo-hydrolase, and PbgA appears to have evolved to support an unknown essential function in E.coli ^[1].

Figure 9: PbgA structural alignments and molecular dynamics simulations. Structure-based alignment of PbgA and EptA isolated PDs (left) and TMDs (right).

Figure 10: Electron density in periplasmic IFD domain of PbgA. a) Bilobal electron density observed along the periplasmic membrane leaflet against the IFD of PbgA after PbgA crystallization using 2.0 Å and 4.6 Å //Fo − Fc// maps. b) Modelled cardiolipin does not refine well into the bilobal electron density. c) Modelled lipid A of LPS refines well into the bilobal electron density.

The crystallized and modelled PbgA was further structurally observed using the Fo − Fc maps to figure out what its unknown function could be. Through this process, they were able to observe a strong bilobal electron density along the periplasmic membrane leaflet against the IFD of PbgA (figure 10a) ^[1].

They found that the modelled cardiolipin (extracted and modelled in the same manner as before) did not refine well into the electron density along the periplasmic membrane leaflet against the IFD of PbgA (figure 10b) ^[1].

Since cardiolipin was not responsible for the electron density, two assays were performed:

1. Limulus amebocyte lysate (LAL) Assay
2. MALDI-TOF mass spectrometry

The LAL assay was carried out to measure the LPS content of purified proteins: PbgA, MsbA, MsbA₂₉₃, and Lnt. Lnt is an inner membrane protein that is not known or expected to bind to or transport LPS. PbgA was purified in LMNG detergent as described earlier. E.coli MsbA and E.coli Lnt were cloned, expressed, and purified in LMNG detergent using the same method as for PbgA. For the LPS-free MsbA, MsbA₂₉₃, E.coli MsbA was cloned into a pRK vector and transfected into human embryonic kidney cells. MsbA₂₉₃ was expressed and purified in LMNG detergent using the same method as the other proteins. The LAL chromogenic endotoxin quantification assay was carried out according to the Pierce manufacturer’s instructions to quantify the LPS content in the purified proteins and the results were graphed. They found a high content of LPS in PbgA which suggests that LPS may be responsible for the electron density (figure 11) ^[1].

Figure 11: LPS quantification in different purified proteins. A high content, ~1.6 EU/mL, of LPS is quantified in purified PbgA. ~0.8 EU/mL of LPS quantified in the purified LPS transport protein, MsbA, for comparison. ~0.1 EU/mL of LPS is quantified in purified Lnt, an inner membrane protein that is not known or expected to bind to or transport LPS, for comparison. No LPS content was quantified in the LPS-free MsbA, MsbA293, for comparison.

The MALDI-TOF mass spectrometry assay was performed to detect various lipid A species present in purified PbgA and Lnt. Kdo₂-lipid A was extracted from purified PbgA and detected using the Bligh-Dyer protocol. The resulting pellet was dissolved in methanol-chloroform and the sample was added to the matrix of the MALDI-TOF/TOF Analyzer. The MALDI-TOF analysis produced a spectrum that was used to identify the lipid A species present in purified PbgA and Lnt.

The following lipid A species were found in purified PbgA (figure 12):

• Tetra-acylated monophosphoryl lipid A
• Tetra-acylated lipid A
• Hexa-acylated lipid A
• Hexa-acylated lipid A + N-arabinose ^[1].

No lipid A species were detected when the procedure was repeated with purified Lnt protein ^[1].

Figure 12: Lipid A species detected in purified PbgA and Lnt through MALDI-TOF mass spectroscopy. Various lipid A species were detected in purified PbgA (black line). The lipid A species, in reducing intensity, are tetra-acylated monophosphoryl lipid A at 1324 m/z, hexa-acylated lipid A at 1797 m/z, hexa-acylated lipid A + N-arabinose at 1949 m/z, and tetra-acylated lipid A at 1430 m/z. No lipid A species were detected in purified Lnt (yellow line).

These assays identified LPS’s lipid A presence in the electron density along the periplasmic membrane leaflet against the IFD of PbgA (figure 13) ^[1]. Lipid A was then modelled and found to refine well into the electron density along the periplasmic membrane leaflet against the IFD of PbgA. This confirmed its presence in the bilobal electron density (figure 10c, 13) ^[1].

Figure 13: Co-purified LPS binds to PbgA’s α7 helix. The presence of lipid A of LPS detected in the electron density along the periplasmic membrane leaflet against the IFD of PbgA via Limulus amebocyte lysate (LAL) assay and MALDI-TOF mass spectroscopy.

Based on this, they concluded a co-purifying LPS molecule remains bound to PbgA via its lipid A part ^[1]. By performing conversation analysis generating structural figures, they found that PbgA’s IFD is responsible for the coordination using a highly conserved periplasmic lipid-A binding motif (figure 14) ^[1].

Figure 14: PbgA’s structural features. a) PbgA crystal structure in cartoon (left) and electrostatic (right) representation. The TMD, IFD and PD are shown in blue, pink, and green, respectively, with LPS as green sticks. b) TMD-based alignment with the PbgA–IFD in pink.

Further structural analysis and sequence analysis was performed to understand the interaction between PbgA and LPS. Sequence analysis was performed using an in-house bioinformatics pipeline that used R and Bioconductor packages, GenomicRanges, GenomicAlignments, VariantTools, and gmapR. Structural analysis was performed using PYMOL.

Figure 15: α7 helical periplasmic LPS-binding motif of PbgA. a) A close-up view of the lipid A-binding motif (pink representation) with LPS (green stick representation), water molecules (blue spheres), and the main bonding interactions (yellow dashes). b) A side view of a.

Figure 16: Structural and sequence characterization of PbgA homologues. a) A close-up view showing the interaction of the Arg215 side chain with a conserved acidic residue (Asp192) in the transmembrane segment (TM5) of the TMD, which stabilizes the IFD–TMD interface. b) Sequence alignment of PbgA sequences from ten //Enterobacteriaceae// gram-negative bacteria. The domain boundaries based on //E.coli//’s PbgA structure are indicated, including the lipid A-binding motif (red shade) and pseudo-hydrolase active site residues (orange triangles).

The structural and sequence analyses revealed that PbgA recognizes a single phospho-GlcNAc unit of lipid A using eight residues that precede and form part of PbgA’s α7-helix, 210-YPMTARRF-217 (figure 15a) ^[1]. PbgA interacts with the unit of lipid A primarily through ten backbone- and water-mediated interactions in the 14-point interaction network:

• Phe-217 anchors the α7-helix within the membrane. Its backbone bonds to the R-3-hydroxymyristoyl and 1-phospho-GlcNAc of lipid A, both through water (figure 15) ^[1].
• Amides of Arg215 and Arg216 complex with 1-phospho-GlcNAc. This is further stabilized by the α7-helical dipole (figure 15) ^[1].
• The Arg215 side chain interacts with an acidic residue, Asp192, that is found in transmembrane segment 5 (TM5) of the TMD and is conserved across all PbgA homologues (figure 16) ^[1]. This interaction appears to stabilize the IFD-TMD interface.
• Thr213 backbone engages with the 3-linked R-3-hydroxymyristoyl group. This interaction is possible because Ala214 links to the 210-YPMT-213 segment (figure 15a) ^[1].
• Thr213’s hydroxyl interacts with the 1-hydroxyl position and 1-phospho-GlcNAc of lipid A (figure 15a) ^[1].
• Met212 wedges between the 2- and 3-linked R-3-hydroxymyristoyl groups and forms hydrophobic contacts (figure 15) ^[1].
• The backbone of Pro211 binds to the 3-linked R-3-hydroxymyristoyl through water (figure 15a) ^[1].
• The backbone Tyr210 binds to the 3-linked R-3-hydroxymyristoyl (figure 15a) ^[1].

To investigate the role of the LPS-PbgA interface in the outer membrane, Clairfeuille et al. (2020) introduced point mutation in the lipid A-binding motif and measured the outer membrane integrity. They replaced Met212 with a charged amino acid (such as glutamate) which compromised the outer membrane by evaluating IC₅₀, whereas mutation to alanine showed a similar level of IC₅₀ as the wild-type ^[1].

The point mutation of T213V did not show strong evidence to prove it affects rifampicin sensitivity ^[1]. However, introducing a T213D mutation made the outer membrane highly sensitive to rifampicin, which was meant to prevent the interaction between PbgA and the 1-phospho-GlcNAc ^[1]. R216A did not show an effect on the outer membrane integrity, but acidic mutations repelled the 1-phospho-GlcNAc group, which lead to rifampicin sensitivity ^[1]. They introduced multiple mutations (M212A/T213V/R216A) at the same time that only simply produced a modest effect, which showed that the multipoint backbone-mediated coordination scheme is important in PbgA ^[1].

Those findings indicate that only mutations resulting in disrupting lipid A-binding function have a significant impact on the outer membrane integrity and concluded that the LPS-PbgA interface is crucial for maintaining outer membrane homeostasis in E.coli.

Clairfeuille et al. hypothesized that a peptide derived from the interfacial domain (IFD) sequence might bind to LPS in vitro. Thus, they used two synthetic, linear peptides that were predicted to bind to LPS and two that were expected to destabilize key lipid A-binding determinants as they all contained the lipid A-binding (LAB) motif from PbgA. These peptides are LABWT, LABWT+, LABΔα7, and LABT213D.

Figure 17: Biolayer interferometry (BLI) data of LAB peptides binding to LPS along with its respective peptide sequence and Kd values.

The LABWT peptide has the same LAB sequence as wild-type PbgA. LABWT+ differs from LABWT by two-point mutations, H221W and D225R. These amino acids were chosen because they are predicted to not have any important interactions between PbgA, and lipid A mentioned in the previous results above. The substitution of H to W, a positive to a hydrophobic amino acid, and the substitution of D to R, a negative to positive amino acid is intended to promote membrane partitioning so the peptides can interact more and insert themselves within the hydrophobic and negatively charged cell membrane. In the peptide LABΔα7, the α7 helix typically in PbgA is deleted, which is an important part of the 14-point interaction network between PbgA and lipid A. In peptide LABT213D, it contains the mutations mentioned in the previous result which showed that this point mutation prevented the interaction between PbgA and the 1-phospho-GlcNAc in lipid A.

Clairfeuille et al. found that LABWT bonded to LPS selectively over all major E.coli phospholipids ^[1]. The LABΔα7 and LABT213D peptides did not bind to LPS ^[1]. The LABWT+ peptide had improved affinity for LPS while having selectivity over phospholipids ^[1].

Biolayer interferometry (BLI), an optical method for evaluating the kinetics and affinities of macromolecular interactions by examining interference _D values. A BLI experiment involves immobilizing one molecule on a Dip and Read Biosensor and measuring its binding to a second molecule. A change in the number of molecules linked to the biosensor tip results in a shift in the interference pattern, which is monitored in real-time. Binding to phospholipids and Kdo₂-lipid A was evaluated at doses of 150, 100, 50, 25, and 10 μM with 300 s association and dissociation steps using biotinylated-LAB peptides. Dissociation constants for LABWT and LABWT+ interactions with Kdo₂-lipid A were estimated by plotting response values at equilibrium as a function of concentration.

LABWT had a K_D value of approximately 75 μM, while LABWT+ had a K_D value of approximately 55 μM, which is smaller than LABWT’s K_D value ^[1]. This means that the point mutations that were intended to increase membrane partitioning also allowed for a greater binding affinity of the ligand for its target.

Clairfeuille et al. tested the PbgA-derived synthetic LAB peptides for antibacterial activity by measuring minimal inhibitory concentrations (MICs) to determine whether it would inhibit the growth of diverse Gram-negative bacteria, including polymyxin-resistant strains. They performed MIC and time-kill assays, which are used to analyze antimicrobial agents’ activity against a bacterial strain to determine the bactericidal activity of an agent over time.

For the MIC, LAB peptides, provided by Smartox Biotechnology, were diluted in LB. For the potentiation MIC assay, log-phase cultures were diluted to an OD600 of 0.0002 in a final volume of 50 μL with peptides and antibiotics. Then, it was incubated overnight at 37°C before growth was measured using a SpectraMax M5 plate reader.

Three separate cultures of wild-type E.coli (ATCC 25922) were grown to log phase before being diluted into varied concentrations of peptide polymyxin B and incubated at 37 °C for the time-kill assay. At various stages, samples were collected, diluted in PBS, and plated on LB agar. After overnight incubation, CFUs were counted.

They found that the LABWT peptide did not affect E.coli growth, which may be because of its large molecular mass of greater than 2 kDa, meaning it was too large to penetrate the bacteria ^[1]. The LABWT+ peptide had MICs of 25–400 μM in chemically or genetically permeabilized cells ^[1]. LABΔα7 and LABT213D peptides showed no effect on cell growth under the same conditions ^[1]. Alanine-scanning, which is a site-directed mutagenesis technique used to determine the contribution of a specific residue, and truncation studies produced a synthetic peptide (LABv2.0) with a MIC of 200 μM against intact, wild-type E.coli K-12.

Clairfeuille et al. further modified lipid A-binding peptide (LABv2.1) that introduced a salt bridge bound to 1-phospho-GlcNAc and an A214F mutation ^[1]. These changes were introduced to improve hydrophobic interactions with LPS and the membrane. Through measurement of minimal inhibitory concentrations and analysis of LPS-PbgA structure, researchers found that LABv2.1 peptide exhibited strong activity against clinically relevant pathogens Enterobacter cloacae, Klebsiella pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa. LABv2.1 demonstrated little effect on the gram-positive bacterium S.aureus that lacks LPS ^[1]. Implementation of PMX-resistance determinants did not change the MIC values (table 1) ^[1]. Therefore, LAB peptides and PbgA can bind to unmodified and modified LPS ^[1].

Table 1: Strains of //E.coli// and their associated minimal inhibitory concentrations of LABv2.1 peptides.

LABv2.1 peptide exhibited bactericidal activity with a different time-kill pattern compared to PMX antibiotics, and it enhanced the effectiveness of outer membrane-impermeable antibiotics while working in synergy with PMX-E.

Red blood cell lysis assay was used to determine the toxicity and cytocompatibility of treatments when applied to human cells (Evans et al., 2013). This assay involves a PBS negative control and a Triton X-100 positive control. The positive control assumes 100% cell lysis as Triton X-100 destabilizes erythrocyte membranes. LABv2.1 and polymyxin E demonstrate lower optical density, correlating to reduced RBC lysis (figure 18) ^[1]. Analogues of LABv2.1 had higher levels of RBC lysis which demonstrated limited specificity to lipid A in gram-negative bacteria ^[1]. Wild-type LAB peptides demonstrated almost no RBC lysis while LABv2.1 demonstrated limited RBC lysis at lower concentrations which increased with a higher concentration of the peptide ^[1]. These findings reveal the discovery of a new class of selective lipid A-binding peptides that can combat gram-negative pathogens, including those that are resistant to PMX modifications.

Figure 18: Left side shows the optical density of samples in a red blood cell lysis including LABv2.1 and polymyxin E. Right side demonstrates % RBC lysis in wild-type LAB peptides, LABv2.0, LABv2.1, and LABv2.1 analogues.

Figure 19: Mass spectrometry analyses of co-immunoprecipitation experiments.

To determine what else PbgA interacts with in E. Coli and where, Clairfeuille et al. performed an immunoprecipitation experiment. The specific type was a co-immunoprecipitation experiment, which is the precipitation of intact protein complexes. PbgA was the known target protein and the antigen. Using an antibody to PbgA, they were able to isolate PbgA from E. Coli, along with any other proteins bound to PbgA. The complexes eluted from this experiment were then analyzed using mass-spectrometry to identify the unknown proteins bound to PbgA. The mass spectrometry results were analyzed using the Significance Analysis of INTeractome (SAINT) algorithm, which calculates the probability of interaction between two components. SAINT scores were assigned from 0-1, with 1 being the highest probability of true interaction. As seen in Figure 19, the darker the blue, the greater the probability of interaction. The sizes of the circles represent their abundance in a specific region. They found that PbgA interacts with the IM proteins LapB (0.67) and PlsY (1).

LapB forms a complex with the protease FtsH in the IM to degrade LpxC, the precursor to LPS. Previous studies had shown that PbgA doesn't interact with FtsH, and the team verified this via a bacterial two hybrid system. This method was used to determine whether there were any protein-protein interactions. As seen in Figure 20, blue colonies form if protein-protein interactions exist, and white colonies form if proteins do not interact with PbgA. This indicates that PbgA interacts with the LapB-FtsH complex through LapB.

Figure 20: Bacterial two hybrid system results. PbgA interactions with LapB bait (blue). No PbgA-FtsH interactions (white).

Researchers concluded that PbgA interacts with LapB and knowing that the LapB-FtsH complex degrades LpxC, the researchers then wanted to determine how PbgA's interaction with LapB affects LpxC downstream. To do this, they performed a Western blot experiment. As seen in Figure 21, they found that when PbgA was overexpressed, LpxC levels increased, and when PbgA was deleted, LpxC was not present. This indicates that PbgA inhibits the LapB-FtsH complex from degrading LpxC.

Figure 21: Western blot results. Overexpression of PbgA leads to increased LpxC, deletion of PbgA leads to absence of LpxC.

As seen in Figure 22, PbgA's position in the periplasmic leaflet allows it to detect any LPS present in the periplasm. When no LPS is detected in the periplasm, PbgA inhibits the LapB-FtsH complex by directly binding LapB. This prevents the degradation of LpxC, and allows LPS biogenesis to proceed to meet the OM demand for LPS. When there is excess LPS at the OM, accumulated periplasmic LPS, awaiting transport to the OM, will be detected by and bound to PbgA. This prevents inhibition of the LapB-FtsH complex by PbgA, as PbgA can no longer bind LapB, and allows LpxC degradation to occur, to slow LPS biogenesis.

Figure 22: Model for PbgA control of LPS biogenesis.

Before the paper by Clairfeuille et al. was published in 2020, there were previous studies done that suggested that PbgA might bind to cardiolipin. However, this study disproves that theory because cardiolipin does not co-purify with PbgA, does not bind to the IFD, and is not necessary to maintain outer membrane integrity in E.coli. After determining high-resolution crystallographic data, this paper shows that PbgA binds to lipid A, not cardiolipin. In addition, before this paper, the mechanism of how LPS synthesis and transport are coordinated to preserve outer membrane integrity has always been unclear. However, in this paper, the structural basis of the crucial LPS-PbgA interaction within the inner membrane was discovered and provides additional insight into the mechanism.

This paper also discovered how unique PbgA is, as it does not seem to require divalent cations or basic residues to bind to lipid A, which is unusual for selective lipid recognition. In addition, it only targets a single phospho-GlcNAc unit on LPS. This differentiates PbgA from other known LPS receptors, LPS transporters, and outer membrane proteins that interact with lipid A. Overall, this publication made many novel discoveries including characterizing PbgA as a key and essential regulator in LPS biogenesis and outer membrane integrity. The researchers also determined the mechanism by which PbgA directly detects LPS on the periplasmic leaflet of the inner membrane.

However, the researchers still have some unanswered questions after their research for this paper. For example, it is still unknown how LPS binding alters the LapB-PbgA interaction and modulation of FtsH activity. An inability to sense LPS results in altered LPS levels and causes a defect in the outer membrane which only exists in the PbgA-TMD strain, but it is still unclear as to why this mutant remains viable.

Furthermore, as polymyxin resistance is becoming more common, other strategies will have to be investigated. This paper shows that the interaction between LPS and PbgA can be a compelling antibacterial strategy since it can be used to disrupt the periplasmic LPS-PbgA interaction. The research done presents opportunities for future antibiotic discovery, since the PbgA peptide can overcome the modifications that cause polymyxin resistance. It was found that this peptide can kill Gram-negative bacteria in vitro, even those strains resistant to polymyxins.

There needs to be more research done for an antibiotic to be made and tested. A paper published in January 2023 discussing the chemical basis of combination therapy to combat antibiotic resistance mentions PbgA ^[2]. They cite the Clairfeuille et al. paper and the potential of this protein to be used as a new antimicrobial agent.

1. Clairfeuille T, Buchholz KR, Li Q, Verschueren E, Liu P, Sangaraju D, et al. Structure of the essential inner membrane lipopolysaccharide–PbgA complex. Nature [Internet]. 2020 Aug [cited 2023 Feb 1];584(7821):479–83. Available from: http://www.nature.com/articles/s41586-020-2597-x

2. Si, Z., Pethe, K., & Chan-Park, M. B. (2023). Chemical basis of combination therapy to combat antibiotic resistance. JACS Au. https://doi.org/10.1021/jacsau.2c00532