Outer Membrane–Peptidoglycan Anchoring in Pseudomonas aeruginosa

Abstract

Gram-negative bacteria have an elaborate envelope that is composed of an outer membrane, the cell wall, and the inner membrane, which collectively encase the cytoplasm. The outer membrane is covalently anchored to the peptidoglycan, the major constituent of the cell wall. We document that the gene product of PA2854 is the catalyst that performs this transformation between the cell wall and the outer-membrane lipoprotein OprI in live Pseudomonas aeruginosa. Furthermore, we reconstitute this reaction in vitro with the use of purified recombinant PA2854, OprI, and synthetic samples of the cell-wall peptidoglycan, in each case documenting the attachment of the side-chain ε-amino group of Lys83 of OprI to the peptide stem of the peptidoglycan. OprI forms a trimeric structure. The X-ray structure of the trimeric state of OprI was solved to 2.1 Å resolution, which reveals it as an extended 82 Å helix bundle. The X-ray structure of PA2854 was also solved at 2.63 Å resolution. The enzyme is composed of three domains, which were documented to bind to the synthetic peptidoglycan and to OprI. The enzyme turns over both non-cross-linked and cross-linked peptidoglycan as its substrate. A model for the ternary assembly of the complex of PA2854-OprI-peptidoglycan is proposed based on the collective evidence. We document that anchoring of the outer membrane to the cell wall catalyzed by PA2854 does not appear to be a redundant reaction (product of more than one enzyme), in the absence of which the bacterium shows a weakened envelope, prone to disruption.

Graphical Abstract

INTRODUCTION

Peptidoglycan (PG), the major constituent of the bacterial cell wall, is composed of a repeating disaccharide unit of β-(1 → 4)-N-acetylglucosamine-N-acetylmuramic acid (abbreviated as GlcNAc-MurNAc). A peptide stem is linked to the N-acetylmuramic acid, which is composed of a uniquely bacterial pentapeptide at full-length: l-¹Ala-γ-d-²Glu-m-³DAP-d-⁴Ala-d-⁵Ala (m-DAP for meso-diaminopimelate) in Gram-negative bacteria. The peptide stem is the site of enzymatic cross-linking of two neighboring strands of peptidoglycan with linkages of both 3,3 (minor) and 3,4 (major) variety. The cross-linking transfigures the cell wall into a large polymer that encases the entire bacterium. The sugar backbone and the peptide stem are modified by a series of periplasmic-protein complexes of Pseudomonas aeruginosa involved in biosynthesis, recycling, and remodeling of the cell wall, which are of interest to our laboratories.^1-5 The gene product of PA2854,² whose sequence analysis argues that it is an l,d-transpeptidase,^6-8 has emerged from these studies. Although l,d-transpeptidases are associated with the formation of the 3,3-cross-links in the bacterial cell wall, they have been shown to incorporate noncanonical d-amino acids into the cell wall⁹ and attach outer-membrane proteins to the cell wall.^10-12 This outer membrane-PG anchoring is important for maintaining the structural integrity of the Gram-negative cell envelope, ensuring mechanical stability under stress and preserving the outer-membrane function as a permeability barrier against harmful agents, including antibiotics.¹³

We report herein how PA2854 performs cross-linking of OprI (an abundant outer-membrane lipoprotein encoded by the PA2853 gene) and the cell wall in live P. aeruginosa (Figure 1). We document that in the absence of the function of PA2854 in a transposon-insertion mutant strain, the bacterial envelope is compromised as a barrier. Reconstitution of this event was recapitulated in vitro with purified recombinant proteins and synthetic peptidoglycan fragments. Furthermore, we present the X-ray structure of PA2854—the first structure of an enzyme involved in this process—together with the X-ray structure of the substrate OprI, providing the opportunity to disclose a model of how this important reaction takes place in bacteria.

Figure 1.

Catalytic process of PA2854. PA2854 (orange) catalyzes cross-linking of the peptidoglycan to OprI (green) in the periplasmic space of P. aeruginosa. The reaction involves a thioester intermediate, which entraps the side-chain amine of the C-terminal lysine of OprI (final residue of OprI-EKASRK shown in green circles).

RESULTS AND DISCUSSION

Characterization of the Function of the Gene Product of PA2854

We studied the functional role of PA2854 in performing the chemical anchoring of the outer membrane by comparing the PG composition of the wild-type P. aeruginosa MPAO1 and that of the transposon mutant MPAO1 PA2854::Tn, which we obtained from the Manoil library (Figure 2).¹⁴ The sacculus, i.e., the purified intact cell wall, was prepared from each strain. As depicted in Figure 2A, processing of the wild-type OprI-linked cell wall by the V8 protease (from Staphylococcus aureus V8 strain; red arrow) leaves a remnant of the C-terminus of OprI (the peptide sequence KASRK) linked to the cell wall, which served for the diagnostic identification by mass spectrometry. Subsequently, the sacculus was digested with mutanolysin (purple arrows) and the reducing ends of the sugars were reduced with NaBH₄. The sequential reactions of mutanolysin and borohydride produced fragments that could readily be identified by LC/MS analysis.

Figure 2.

Detection of Tri-KASRK muropeptides in sacculus from the wild-type strain and their absence in the sacculus of the PA2854::Tn mutant strain. (A) The strategy to generate OprI-linked muropeptides for LC/MS analysis using the V8 protease and mutanolysin is depicted. The red and purple arrows mark the cleavage sites of the V8 protease and of mutanolysin, respectively. The measured m/z values for nonlipoprotein-linked products are shown, while theoretical values are included in parentheses. LC/MS traces of muropeptides from (B) the wild-type and (C) the PA2854::Tn strains are shown. Top traces are TIC, and bottom traces are EIC of OprI-linked muropeptides (1–3). (D) Mass spectra of peaks 1–3 confirming their masses are consistent with the presence of Tri-KASRK. (E) CID mass spectrum of precursor 1, highlights fragment ions that confirm the m-DAP-ε-Lys cross-link with the corresponding Tri-KRSAK structure shown in panel (A). The 600–1200 Da range of the mass spectrum is expanded to show smaller-sized peaks. Cleavage of GlcNAc–MurNAc_r result in ions B1 and Y1, while ions resulting from amide-bond cleavage along the tripeptide stem or the KASRK are labeled b1–b4 and y1–y4 and colored in black (tripeptide) or green (KASRK). Ions that retain the m-DAP-ε-K linkage are labeled in bold, while underlined labels mark fragments that have lost GlcNAc. G signifies GlcNAc, and MurNAc_r reduced MurNAc; A is for alanine, E is the γ-d-²Glu, and m-DAP m-diaminopimelate. The presence of an asterisk (*) denotes the replacement of MurNAc_r by anhMurNAc.

As shown for the sacculi prepared from the wild-type (Figure 2B) and the transposon mutant (Figure 2C) strains, the total-ion-chromatogram (TIC) traces display the expected abundant PG monomer and dimer species (Tetra, TetraTetra, TetraTetra*). In this nomenclature, Tetra and Tri refer to N-acetylglucosamine-β-1,4-N-acetylmuramic acid with a tetra- or tripeptide stem (GlcNAc-MurNAc-Tetra/Tri). Similarly, TetraTetra represents two Tetra units joined by a peptide cross-link, and TetraTetra* contains one disaccharide borohydride-reduced MurNAc (MurNAc_r) replaced by 1,6-anhydro-N-acetylmuramic acid (anhMurNAc; found at the termini of the individual strands of PG), as depicted in Figures 2A and S1. Furthermore, the extracted-ion chromatograms (EICs) of the samples from the wild-type MPAO1 show three additional peaks corresponding to OprI-linked muropeptides 1–3 (Tri-KASRK, TetraTri-KASRK, and TetraTri-KASRK*; the underline K designates the lysine attached to the peptidoglycan at its side chain), which were entirely absent from the MPAO1 PA2854::Tn mutant samples. The absence of OprI-linked PG in the sacculus of the mutant strain reveals that PA2854 cross-links OprI to the peptidoglycan. OprI-linked muropeptides from the wild-type sacculus show both non-cross-linked (e.g., 1) and cross-linked (e.g., 2 and 3) PG species. The important conclusions are that OprI can be attached to both the cross-linked and non-cross-linked stem peptides and that there is no apparent redundancy in the P. aeruginosa genome for the function of PA2854. If indeed there exists another enzyme in P. aeruginosa that performs this reaction, that of PA2854 dominates.

High-resolution accurate-mass data for compounds 1–3 agreed with the calculated masses for the proposed structures (Figures 2D and S1). Compound 1 was further analyzed by collision-induced dissociation (CID) mass for the purpose of confirmation of the assignments, which produced the expected series of b/y ions from peptide-bond cleavage and B/Y ions from glycosidic-bond cleavage, as annotated in Figures 2E and S1C. The fragment ions highlighted in bold (Y1, y1, y2, y3, y4) all retain the DAP–ε–Lys linkage, confirming that the KASRK peptide is covalently attached to the peptidoglycan stem.

Consequences of the In Vivo Loss of Functional PA2854

The functional loss of the PG-outer-membrane anchoring is not lethal, as evidenced by the growth of MPAO1 PA2854::Tn. One consequence of losing PG-outer-membrane anchoring is a more fragile outer membrane.¹³ With our biochemical positioning of PA2854 as the sole catalyst for PG anchoring to the outer membrane protein OprI, we sought to investigate its role in preserving outer membrane integrity. We conducted scanning electron microscopy experiments to compare the morphological changes of strain MPAO1 under stress in the presence and absence of PA2854 function. We exposed log-phase wild-type MPAO1 and MPAO1 PA2854::Tn to either isotonic or hypotonic conditions. The hypotonic shock was induced by brief exposure to 1 mM EDTA in water. Addition of EDTA depletes divalent cations, which normally coordinate lipid A to stabilize the outer membrane; therefore, the differences in morphology between the two strains are clearer. The cells were then imaged by scanning electron microscopy.

Both the wild-type (Figure 3A) and mutant (Figure 3B) bacteria exhibited similar morphology under isotonic conditions. The cell shape, size, and surface structure did not change with the loss of the PA2854 activity. However, we observed some disrupted bacteria in the case of the mutant strain (Figure 3B). A more drastic difference was observed under divalent ion depletion conditions in the presence of EDTA. The wild-type strain showed intact bacteria but also some that appeared deflated as they had burst under hypotonic conditions (Figure 3C). The mutant strain displayed multiple prominent protrusions on the surface with a spider-web appearance due to leaky membranes (Figure 3D). The cross-linking of the PG to OprI by PA2854 is not critical for bacterial survival but would appear to play an important role in maintaining the outer-membrane stability/integrity.¹³

Figure 3.

Scanning electron microscopy of P. aeruginosa. Images show (A) the wild-type MPAO1 and (B) PA2854::Tn MPAO1 strains under isotonic conditions in phosphate-buffered saline. (C) The wild-type and (D) the mutant strains were treated with 1 mM EDTA. All images are at 50,000× magnification. White bar is 2 μm.

OprI Exists in a Monomer-Trimer Equilibrium

Having documented the reaction of PA2854 in live P. aeruginosa, we turned our attention to reconstituting the reaction in vitro with recombinant purified proteins and synthetic PG samples. We expressed the coding sequences and purified recombinant PA2854 and OprI to homogeneity (Figure S2A). We confirmed the molecular mass of PA2854 by mass spectrometry (Figure S2C). In the case of OprI, mass spectrometry under nondenaturing conditions showed both monomeric (6504 Da) and trimeric (19,513 Da) species (Figure S2B). Significant trimerization was seen at concentrations of OprI exceeding 20 μM, which allowed for estimation of a K_D at ~20 μM. No dimeric species were detected in our mass spectrometry experiment, indicating that trimerization occurs over a single kinetic transition. OprI accounts for approximately 20% of the outer-membrane proteins in P. aeruginosa.¹⁵

PA2854 Interacts with Its Substrates In Vitro

Next, we attempted thermal-shift assays to demonstrate that indeed PA2854 interacts with its substrates in vitro. This experiment was important for the synthetic PG samples, since the actual PG, a polymer, is being approximated by the synthetic samples. For this assessment, we prepared the synthetic PG samples 6 and 7 (structures of compounds 4-7 are given in Figure S3A), each with a tetrapeptide stem, by methods that we have reported previously.^16-19 The native OprI is a lipoprotein; however, the anchor for the membrane is absent in the recombinant sample. These experiments were possible in the light of the presence of multiple tyrosines and phenylalanines in PA2854 and only one tyrosine per OprI monomer. This gave a fluorescence signature with reliable inflection temperatures for PA2854, and a total absence of fluorescence for OprI. We observed a clear temperature shift for PA2854 when it was incubated with the synthetic GlcNAc-anhMurNAc-Tetra (6), GlcNAc-MurNAc-Tetra (7), OprI, or OprI plus either 6 or 7 (Figure 4A). The “Tetra” in these compounds is for tetrapeptide as the stem. These experiments indicated that recombinant PA2854 interacts with both the synthetic peptidoglycan substrate(s) and OprI in vitro.

Figure 4.

Reconstituting the reaction of PA2854 in vitro. (A) Melting temperatures of PA2854 and its complexes were measured by nanoDSF. All measured ΔT_i values are statistically significant (n = 3, p < 0.0001 for all values, p = 0.0002 for PA2854 with OprI and 6). (B) Reactions of synthetic PG tetrapeptides 4–7 with OprI catalyzed by PA2854 are shown; in the absence of the enzyme, no reaction took place. The deconvoluted mass spectra of OprI and of the ligated products OprI–PG demonstrate the formation of products 4p–7p. (C) Reaction of Meropenem (MEM) with PA2854 allowed detection of the PA2854-MEM species. The deconvoluted mass spectra of PA2854 and of the covalent acyl-enzyme complex formed with Meropenem (PA2854-MEM) are shown. The black bar indicates the +383 Da mass shift from the signal for apo-PA2854.

In Vitro Reconstitution of the Catalytic Reaction of PA2854

To document the reaction of PA2854, we incubated the enzyme with OprI and with the aforementioned synthetic PG samples individually (compounds 4-7; Figures 4B and S3). We used non-nucleophilic HEPES buffer and performed all reactions at 4 °C. Products for the coupling of substrates 4–7 with OprI, compounds 4p–7p, respectively, were detected by intact-mass analysis; the quantities of each species incidentally increased as a function of time (Figures 4B and S4A). The peak at 6504 Da corresponds to unmodified OprI, whereas the additional peaks (4p–7p; “p” denotes product) represent OprI modified by a single PG sample with a tripeptide stem (Tri–OprI). The difference in mass of each product reflects the mass of the corresponding appended PG, as illustrated in Figure 4B. The transformation clearly accounts for the ligation of OprI to the synthetic PG sample(s) in a process that uses the tetrapeptide stem in the substrate PG for the linkage to the side chain of lysine in OprI. This ligation of the two substrates entails the departure of the terminal d-Ala from the tetrapeptide in the PG substrate(s), which is accounted for in the mass spectrometry data.

Cross-linked PG substrates containing a TetraTetra stem motif were also tested to determine whether TetraTri–OprI adducts would result (Figure S3). These substrates were generated by digestion of the sacculus with the lytic transglycosylase MltA, which produced GlcNAc-anhMurNAc-containing species 6 (Figure 4B) and TetraTetra 8 (Figure S3) as the two most abundant species, or with lysozyme, which produced GlcNAc-MurNAc-containing species 7 (Figure 4B) and TetraTetra 9 (Figure S3). The resultant mixtures were subjected to the PA2854/OprI ligation. Under these conditions, PA2854 performed the ligation reaction (to OprI) with the non-cross-linked donors 6 and 7, but also with the cross-linked donors 8 and 9 (Figure S3B,C).

Covalent Inhibition of PA2854 by Meropenem

Sequence analysis of PA2854 revealed it to have a YkuD domain, found in ld-transpeptidases. ld-Transpeptidases are typically inhibited by carbapenems.²⁰ Under the conditions that we documented the reaction of PA2854, we also observed the formation of the covalent PA2854–Meropenem adduct, which increased with longer incubation time (Figures 4C and S4B). As shown in Figure 4C, apo-PA2854 appears as two closely spaced species corresponding to loss of the N-terminal Met (34,988 Da) and loss of Met plus gluconoylation (35,166 Da). N_α-Gluconoylation on the His-tag N-terminus is a well-known transformation.²¹ Meropenem modifies each form to give a covalent adduct (PA2854–MEM), which is shifted by +383 Da in the deconvoluted mass spectrum (Figure 4C). Time-course analysis under these conditions yielded an observed reaction rate k_obs = 0.066 ± 0.004 h⁻¹ with a half-life of t_1/2 = 10.5 ± 0.6 h (Figure 4C). These kinetic parameters are reflective of a slow acylation process, consistent with the slow cross-linking reaction catalyzed by PA2854 in the presence of its substrates. Despite the slow nature of this reaction, the transformation is nearing quantitative with fraction modified (F_max) at 95.4 ± 1.9%.

X-ray Structure of PA2854

We turned our attention to the determination of the structures of PA2854 and OprI. The sequence of PA2854 is composed of a predicted signal peptide (amino acids 1–23), an N-terminal PG-binding LysM domain (amino acids 48–84), a YkuD catalytic domain (amino acids 96–231),^22,23 and a C-terminal extension (residues 232–323). Structural determination of PA2854 was achieved by the construct that spanned amino acids 24–323 (see the Experimental Procedures Section and Table S1). An N-terminal histidine tag was incorporated, consisting of the sequence MGSSHHHHHHSSGEN-LYFQGHM, which proved essential in preventing protein aggregation upon TEV cleavage. This was consistent with previously reported solubility issues that hindered production of soluble PA2854.⁸ The highest-quality crystals allowed structure determination at 2.63 Å resolution (Figure 5). Notably, the crystals exhibited 72% solvent content—significantly higher than the typical protein (47–50%) among the PDB-deposited structures^24,25—likely contributing to the moderate resolution limit and marked anisotropy. The electron-density map clearly resolved the stretch of amino acids 24–305, whereas no density was observed for the final C-terminal segment (amino acids 306–323) or for the initial 14 amino acids from the His-tag, indicating inherent high flexibility in this C-terminal segment of the protein (amino acids 306–323; Figure S5).

Figure 5.

Three-dimensional structure of PA2854. (A) Domain composition of the PA2854 construct used for crystallographic experiments is shown. The crystal structure (2.63 Å) is shown in solvent-accessible surface overlapped with the ribbon representations, with the YkuD catalytic domain in orange, the LysM domain in purple, and the C-terminal domain (C-t) in dark blue. (B) Close-up view of the YkuD domain in PA2854, highlighting its secondary structure elements and the location of the loop containing the consensus motif (L_C) is given. (C) Structural superposition of the YkuD domain in PA2854 (orange) and that in the Bacillus subtilis YkuD protein (green) shows high similarity. The catalytic Cys207 and His191 and the conserved ArgR209 are depicted in capped stick representation. Main structural differences are labeled. (D) Detailed view of the substrate cap with relevant residues depicted as capped sticks. (E) Zoomed view of the salt-bridge interactions (dashed lines) stabilizing the core of PA2854.

In the crystal structure of PA2854, the N-terminal purification tag of a symmetry-related molecule was inserted into the active site of PA2854 (Figure 7A). The tag establishes different interactions with the PA2854 active site: salt-bridge (Lys180), hydrogen bonds (Trp147, Pro150, Thr182, and Thr139), and van der Waals interactions (Pro150, Leu176, and Leu189). This ensconced tag abrogated all of our soaking experiments with carbapenems or with synthetic peptidoglycan analogues. A truncated version of the protein (amino acids 24–304) bearing the His-tag at the C-terminus was prepared to circumvent this problem, but it failed to crystallize in more than 900 tested conditions. These included trials with antibiotics and seeding from crystals of the full-length constructs. These results underscore the importance of the N-terminal region in promoting crystal formation and protein stability.

Figure 7.

PG binding to PA2854. (A) View of the PA2854 crystal packing reveals insertion of the N-terminal segment of a symmetry-related molecule (gray ribbon) into the catalytic cleft (surface representation, domain-colored). The catalytic cysteine (Cys207) is shown in blue. The N-terminal segment (18–25) of the symmetry molecule occupies the acyl-donor site, with the boxed portion shown in expansion. Key interacting residues from the YkuD catalytic domain are shown as capped sticks; hydrogen bonds are indicated by dashed lines. Catalytic residues Cys207 and His191 are highlighted in bold. (B) Model of the acylated intermediate for PA2854:GlcNAc-MurNAc-l-Ala-γ-d-Glu-m-DAP. Main residues involved in PG stabilization are represented as capped sticks and labeled.

The PA2854 structure revealed a three-domain architecture (Figures 5A and S6A), composed of a LysM domain (amino acids 40–84), a catalytic YkuD domain (amino acids 96–231) and a C-terminal domain (amino acids 235–303). The LysM domain adopts the typical βααβ fold (Figure 5A) that has been reported to recognize the GlcNAc-X-GlcNAc motif in different polysaccharides such as the peptidoglycan.²⁶ The LysM domain of the B. subtilis YkuD (PDB: 4A1J)²⁷ is the closest structural homologue (Figure S6B). The catalytic domain presents a YkuD-like fold²⁷ consisting of a β-sandwich formed by two mixed β-sheets (β7-β10-β4-β5-β6 and β7-β8-β9) and a single α-helix between β9 and β10 strands (Figure 5B). The conserved l,d-transpeptidase fingerprint motif²⁸ (Hxx_14–17(S/T)HGCh(R/N) with h for any hydrophobic amino acid) is present in PA2854 (amino acids 190–207; Figure S7). Cys207, the invariant catalytic residue in other l,d-transpeptidases²⁹ exists in this motif (Figure 5B). His191 is the base that promotes it for the active-site acylation by the substrate. The l,d-transpeptidase motif forms a partially disordered loop in the structure, denoted as L_C (residues 191–207) (Figure 5B).

An Inactive-to-Active Structural Transition of PA2854

Structural homologues of the catalytic domain were identified using DALI server³⁰ and Foldseek,³¹ including l,d-transpeptidases that perform the 3,3-cross-linking of the cell wall in various organisms (Figure S6), the closest one being the B. subtilis YkuD (PDB: 4A1J, RMSD of 0.96 Å for 53 Cα atoms superimposition).²⁷ Other structurally related l,d-transpeptidases are Enterococcus faecium Ldt_fm (PDB: 6FJ1)³² and Escherichia coli YcbB (PDB: 6NTW)³³ (Figure S6). Although the global YkuD fold is conserved, there are important differences in that of PA2854. The catalytic cysteine in PA2854 is not exposed, but buried and somewhat remote from the side-chain imidazole of the catalytic histidine (distance Nε2-Sγ of 6.0 Å versus 4.2 Å in B. subtilis YukD) (Figure 5C). Differences are also observed in loops L_122–133 and Lc surrounding the active-site cavity and in the substrate cap (residues 144–178) (Figure 5C). This cap has been implicated in regulating substrate access and product release in homologues.^33,34 Its conformation in PA2854 is stabilized by a hydrophobic patch (Trp147, Pro149, Pro150, Ile153, Leu164, Pro175, Leu176), and a salt-bridge between His205 (Ldt motif) and Glu156 in the cap (Figure 5D). Notably, this region is proline-rich (10 Pro, >28% of residues), suggesting a structural role and/or a potential interface for protein–protein interactions. The C-terminal domain consists of a β-sheet with three β-strands and two α-helices (Figure 5A). This domain is unique to PA2854, as it was not identified by Prosite,³⁵ Pfam,³⁶ nor found in homologues.

The three domains of the protein establish many polar contacts among themselves. The LysM domain contains an extra β-strand (β1) packed with the β6 of the catalytic domain (Figure 5A). Also, an extensive network of salt-bridge interactions connects the three domains at the protein core (Figure 5E), conferring rigidity to this region, whereas the peripheral regions of the domains exhibit high conformational mobility (Figure S5B).

Comparison of the crystal structure of PA2854 with the model predicted by AlphaFold2³⁷ (Figure S8A) reveals that the overall fold is maintained (RMSD of 0.998 Å, 214 Cα atoms). Changes in the relative positions of the peripheral regions of the three domains are observed (Figure S8B), consistent with the mobility detected in the crystal (Figure S5B). Notably, significant conformational differences are detected in loops proximal to the catalytic center (L_C, L_122–133, and L_249–263). The predicted structure displays a fully competent catalytic dyad, as observed in other enzymes with the YkuD domain, such as B. subtilis (Figure 5C), with the active-site cysteine thiol positioned for promotion by the histidine imidazole. Transition between the two conformations requires the concerted backbone rearrangements of these three loops (L_C, L_122–133, and L_249–263) of up to ~9 Å (Figure S8C). The enzyme therefore appears to exist in two conformations; an inactive one, where there is spatial separation between His191 and the thiol of Cys207, and an active conformation, where these groups are in close proximity. The active conformation shows an active-site architecture poised for catalysis. We reconstituted this transition from the inactive conformation to the active conformation using targeted molecular dynamics (Movie S1). In this simulation, we observed the rearrangement of all three loops over the course of 11 ns of simulation time. These motions both organize the binding surface for OprI and properly poise the active site His191 and Cys207 for catalysis. The thiol of Cys207 and the lone pair of electrons in His191 imidazole sample distances between ~4.5 and 5.0 Å. This simulation argues that the PA2854 structure can transition between two states, one suitable for catalysis and another as a latent form, which might exist for regulatory functions of its catalytic process.

OprI Forms a Triple Helical Structure

A soluble version of OprI (spanning amino acids 25–83) crystallized in the cubic space group P2₁3 with one chain per asymmetric unit (see the Experimental Procedures Section and Table S1). The OprI structure, solved at 2.1 Å resolution, comprises a long α-helix (residues 25–80) (Figure 6A) with the final three C-terminal residues in a more flexible conformation, resulting in a less-defined electron density (Figure S5C,D). OprI organizes as an 82-Å-long α-helical trimeric coiled-coil generating the symmetry of the crystal. The X-ray structure is consistent with the trimeric assembly, revealed by what was observed in solution by mass spectrometry (Figure S2B). The coiled-coil organization of three parallel α-helices varies its cross-sectional arrangement along the structure, with an N-terminal triangular cross-sectional side width of ~11.6 Å to a more compact cross-sectional side width at the C-terminal end (8.7 Å) (Figure 6B-E).

Figure 6.

Three-dimensional structure of OprI. (A) Overall view of the crystal structure of OprI is shown. The OprI chain (green) creates the homotrimer with two symmetry-related chains (A′ and A”, colored in light gray and dark gray respectively). Close-up views of the polar interactions are given among the trimer chains, with the protein shown in the ribbon representation, the side chains in capped sticks, and polar interactions as dashed lines. (B) Top view of the trimer structure is shown, with distances indicated. (C) Schematic indicating the single helical span of the protein. (D, E) Representation of the alanine zipper of OprI is given in side and top views, respectively; alanine residues are shown as yellow spheres in the side view and as sticks in the top view.

The OprI chain contains a high percentage of alanine residues (28.1%; 18 alanines). These alanine residues are all sequestered at the interfaces of the trimer, conforming to what can be construed as an “alanine zipper” (Figure 6D,E), essential for the formation of the coiled-coil. The alanine side chains participate in extensive van der Waals interactions along the zipper length, contributing to its hydrophobic core. In addition, polar interchain interactions are observed all along the length of the structure that contribute to trimer stabilization. In the N-terminal region, Arg27 establishes two salt bridges and a hydrogen bond with amino acids of the adjacent chain (Glu28, Glu35, and Thr32, respectively) (Figure 6A). In the central region (Figure 6A), another interchain salt-bridge interaction is formed between Arg44 and Asp46. In the C-terminal region, two interchain hydrogen bonds are formed between Thr65 and Gln63 and between Arg72 and Asn70 (Figure 6A). As these interactions are repeated three times in the oligomer, the global contribution of both the alanine-zipper and the polar interactions makes a stable trimer. This feature supports the notion that OprI behaves physiologically as a trimer, similarly to the lipoprotein Lpp protein in E. coli,^38,39 with which it shares structural homology (RMSD = 1.36 Å for 158 Cα atoms superimposition) (Figure S9). It is worth mentioning that most of the hydrophobic amino acids in OprI are alanine residues (capped by a Leu at the N-terminus and a Met at the C-terminus) (Figure S9B), in Lpp the stretch of alanine residues is shorter and most of the hydrophobic amino acids are bulkier (Leu, Ile and Val), thus occupying half of the length of the lipoprotein (Figure S9C). OprI and Lpp display comparable electrostatic potential profiles (Figure S9D). Both show a basic N-terminus promoting outer-membrane lipid interactions, a largely acidic central region in the exposed helical trimer, and a basic C-terminal segment containing the Ser–Arg–Lys motif required for peptidoglycan binding.

In Silico Reconstruction of the Outer Membrane–Peptidoglycan Cross-Linking Complex

Structural comparison of PA2854 with related enzymes, such as Meropenem-acylated E. coli YcbB³³ (Figure S6D) and Mycobacterium tuberculosis Ldt_MT2 bound to Meropenem and faropenem^34,40 revealed that their ligands and the N-terminal symmetry mate in PA2854 all occupy the same acyl-donor binding site. In order to further explore the relevance of the observed amino acids in the interaction with the acyl-donor stem of PG, a computational model of the acylated PA2854 intermediate with the peptidoglycan fragment GlcNAc-MurNAc-l-Ala-g-d-Glu-m-DAP was built. This complex was constructed in the model of the AlphaFold prediction of the structure for PA2854 (Figure S8). The model was generated by cofolding the PA2854 sequence with GlcNAc-MurNAc-Tetra using RoseTTAFold All-Atom.⁴¹ The tetrapeptide stem inserts into the active site of PA2854, while the glycan units remain at the entrance. The peptide stem was poised for nucleophilic attack by Cys207 for the formation of the thioester intermediate with the attendant displacement of d-Ala. Lys180 makes a salt-bridge interaction with the carboxylate group of the γ-d-Glu and His204 with the carboxylate of m-DAP, while residues Trp147, Ser152, His191, and Pro186 establish hydrogen-bond interactions with the peptide stem (Figure 7B). The van der Waals interactions by Leu189 and Pro150, among others, contribute to stabilization of the intermediate state. Incidentally, all these amino acids interact with the N-terminal tag in the crystal structure of PA2854 in the inactive state (Figure 7A).

The second step of the transpeptidation reaction, leading to anchoring of PG to OprI, differs from the canonical l,d-transpeptidases, which perform the 3,3-cross-linking reaction. AlphaFold-Multimer extension,⁴² provides a high-confidence model of trimeric OprI in complex with PA2854 (Figures 8A and S10). Complexes were modeled with a 1:3 PA2854:OprI stoichiometry, based on trimeric OprI arrangement observed both in solution and in the crystal structure. The model positioned the C-termini of the OprI trimer near the catalytic cavity of PA2854, specifically in a pocket opposite the acyl-donor site—consistent with the expected location of the acceptor site. Notably, acidic surface patches on PA2854—formed by acidic residues located in the substrate cap (Asp160, Asp162, and Asp173) and in the LysM domain (Glu47, Asp52, and Glu55)—appear to stabilize this interaction (Figure 8B). These acidic features are unique to PA2854 and absent in 3,3-l,d-transpeptidases such as B. subtilis YkuD, E. coli YcbB, E. faecium Ldtfm217 (Figure S11), or the hydrolytic enzymes LdtF or DpaA (data not shown), suggesting a specialized role in outer membrane lipoprotein recognition. Importantly, the computational model for the PA2854:OprI-trimer complex inserts the terminal Lys83 from one of the OprI chains in a very close distance of the PA2854 catalytic dyad (Figure 8C).

Figure 8.

Proposed model of outer-membrane anchoring of the cell wall. (A) Views of the AlphaFold-multimer model of PA2854:OprI interactions are given, with PA2854 shown as a surface representation, with domains colored as in Figure 1. The OprI homotrimer is illustrated in green ribbon presentation, and a peptidoglycan fragment occupying the acyl-donor site is shown in CPK representation in white (at 7 o′clock). Boxes indicate regions that are magnified. (B) Close-up view of the interaction between PA2854 and the C-terminal Arg-Lys of OprI inserted into the catalytic site. PA2854 amino acids near the OprI C-terminus are shown as capped sticks, and hydrogen bonds are indicated with dashed lines. (C) Detailed view of the PA2854 active site, showing Cys207 acylated by a peptidoglycan fragment (MurNAc-l-Ala-γ-d-Glu-m-DAP), represented as gray capped sticks for the carbon atoms. (D) Electrostatic surface potential of PA2854 and the OprI homotrimer at the proposed interaction interface, highlighting a charge complementarity between the two proteins. (E, F) Within 100 ns of simulation time, the ε-NH₂ of Lys83 of OprI samples distances (d) and angles of nucleophilic attack (θ) that favor approach of the amine as the nucleophile to the thioester carbonyl. Panel (F) shows a representation of the trajectory for nucleophilic attack and the definitions of d and θ.

The crystallographic coordinates used to generate the complex of OprI and PA2854 (in complex with GlcNAc-MurNAc-Tri) were aligned over the AlphaFold-Multimer model of PA2854:OprI-trimer (see the Experimental Procedures Section). A POPE membrane section was constructed and covalently anchored to the N-terminal cysteine of all three OprI monomers to mimic the environment of membrane-linked OprI. The dynamics of the full complex (PA2854:Glc-NAc-MurNAc-Tri:OprI-trimer:membrane) was simulated for 100 ns (Movie S2). Throughout the simulation, the C-terminal ends of all three OprI monomers were in contact with PA2854 through electrostatic complementarity between OprI and PA2854 (Figure 8D). The C-terminus of OprI is dominated by basic residues, which are complemented by acidic residues at the entrance of the PA2854 active site. This arrangement allows for penetration of the ε-amine of Lys83 of one of the OprI monomers into the active site. The simulation revealed a stable arrangement of the catalytic components required for OprI-peptidoglycan cross-linking. We observed stabilization of the thioester oxygen of acylated PA2854 in an oxyanion hole formed by hydrogen bonding to the amides of Gly206 and Cys207. The ε-amine of Lys83 of the OprI monomer inserted into the active site frequently sampling conformations and distances to the thioester carbonyl that allow the nucleophilic attack to occur (Figure 8E,F and Movie S2). The distance (d) and the angle of nucleophilic attack (θ) approach ~3 Å and 100°, respectively. Our simulations support the double-displacement mechanism for PA2854. A double-displacement mechanism is one in which the first substrate produces a covalent complex with the enzyme. The second substrate then binds and reacts to produce the final product.

CONCLUSION

OprI, abundant at approximately 20% of the outer-membrane proteins,⁴³ exists as a homotrimer as evidenced by mass spectrometry and X-ray crystallography. The trimeric OprI as a lipoprotein is anchored to the inner leaflet of the Gram-negative outer membrane. The membrane environment imposes mechanical and structural constraints on the assembly and orientation of OprI. We estimate the K_D of trimerization at ~20 μM for the recombinant OprI in solution. However, one might expect a different value in the native membrane environment. The natural abundance of OprI in the outer membrane argues that an evolutionary pressure might not exist for significant enhancement of the K_D value.

Meanwhile, the tetrapeptide stem of the peptidoglycan, the most abundant form of the peptidoglycan during homeostasis,¹⁶ undergoes a reaction with the active site cysteine of PA2854 to generate a thioacyl intermediate. The amide carbonyl of m-³DAP in the tetrapeptide stem sequesters within the active-site oxyanion hole for this transformation, which leads to the nucleophilic attack of the carbonyl by the cysteine side chain with the attendant displacement of the terminal d-⁴Ala (Figure 9, top). The thioester carbonyl is promoted for the second step of the double-displacement mechanism by the same oxyanion hole, which is poised for nucleophilic capture of the side chain of Lys83 from trimeric OprI (Figure 9, bottom). We documented this process in both live bacteria and a reconstituted in vitro system with recombinant proteins and synthetic peptidoglycan surrogates. Incidentally, these experiments revealed that the PA2854 can catalyze the ligation reaction with both the non-cross-linked and cross-linked peptidoglycan.

Figure 9.

Catalytic cycle of PA2854. PA2854 (active-site amino acids shown in orange) ligates PG (shown in black) to OprI (shown in green) by the ping-pong mechanism. The catalytic His191 promotes Cys207 for acylation by the PG. The backbone nitrogen atoms of Cys207 and Gly206 comprise the oxyanion hole to stabilize the tetrahedral species, which leads to the formation of the acyl-enzyme intermediate (at 2 o′clock) concomitant with the release of d-⁴Ala from the substrate as the leaving group. The ε-NH₂ of Lys83 of OprI is promoted by the same catalytic machinery, inclusive of a transient tetrahedral species that is stabilized by the oxyanion hole, en route to the formation of the ligation product that has achieved linkage of the membrane-bound OprI to the polymeric peptidoglycan (at 6 o′clock).

The in vitro-reconstituted system exhibits slow reaction, which is also true for the modification of PA2854 by Meropenem. These slow kinetics hint at the requirements of the native environment of the outer membrane and periplasmic space for proper enzymatic function, which has not been duplicated in vitro. Notwithstanding the slow kinetics, we successfully demonstrated the enzymatic reaction both in vitro and in live bacteria.

We proposed in this report a three-dimensional model of how this assembly is formed. The membrane-anchored OprI binds PA2854 by electrostatic surface interactions that showed stability in dynamics simulations and was supported by nanoDSF measurements. In this arrangement, the catalytic and the LysM domains of PA2854 would be positioned within the PG layer, according to the spatial measurements for the periplasm reported earlier.^44,45 The catalytic domain is inserted into the peptidoglycan layer, while the LysM domain remains on the top layers (Figure S12).

The structure of PA2854 reveals a previously unrecognized domain organization among l,d-transpeptidases, including the C-terminal domain with an unprecedented fold. Notably, PA2854 displays distinct acidic surface patches surrounding the active site that mediate specific interactions with trimeric OprI; such a feature is absent from the structures of l,d-transpeptidases that catalyze canonical 3,3-peptidoglycan cross-linking. Together, these findings provide a structural and mechanistic framework for outer membrane–peptidoglycan tethering, a process essential for Gram-negative envelope integrity and likely conserved across bacterial species.

EXPERIMENTAL PROCEDURES

Peptidoglycan fragments 1–4 were synthesized using methodologies previously developed in our laboratories.^16-19 Trypsin-treated E. coli sacculi were prepared as previously described.⁴⁶ These sacculi were used to prepare compounds 6–9 as detailed in the methods. Wild-type MPAO1 and transposon-insertion mutant—lacking functional PA2854 (PA2854::Tn MPAO1)—strains were purchased from the Manoil laboratory (University of Washington at Seattle).

Assessment of PA2854 Reaction Products from the Sacculi

V8-treated P. aeruginosa sacculi were isolated from the wild-type MPAO1 and PA2854::Tn MPAO1 strains. Bacteria were grown in LB media overnight, harvested by centrifugation, and were frozen at −80 °C. Frozen cell pellets were resuspended in 20 mM sodium phosphate buffer, pH 7.5 (25 mL). This cell suspension was added dropwise to 50 mL hot 8% sodium dodecyl sulfate (SDS) and boiled with stirring for 3 h. The SDS solution was removed by repeated (5×) centrifugation at 36,000 rpm at 30 °C in an F37L −8 × 100 rotor (ThermoScientific), followed by resuspension in 80 mL water. After the final wash, the pellet was resuspended in 10 mL of 10 mM Tris-HCl pH 8.0, supplemented with 2.5 mM MgCl₂ and 0.5 mM CaCl₂. RNase A (4 μL, New England Biolabs) and DNase I (4 μL, 10 mg/mL) were used to digest nucleic acids, and the attached proteins were digested by treatment with 0.125 mg endoproteinase Glu-C (Roche; V8 protease) overnight at 37 °C. The enzymes were denatured in 10 mL 8% SDS at 80–90 °C for 1 h and separated from the sacculus by pelleting at 36,000 rpm. Four successive rounds of pelleting and washing with water were performed to remove the SDS. The sacculus was resuspended in 2.5 mL water and sonicated for 20–30 pulses before being stored at −80 °C. Sacculus was treated with mutanolysin overnight. The solution was centrifuged, and the supernatant was passed through a 10 kDa molecular-weight-cutoff (MWCO) Amicon filter. The filtrate was reduced with NaBH₄ and analyzed by UHPLC/MS method 1, as described below.

Ultraperformance Liquid Chromatography/Mass Spectrometry (UPLC/MS)

UPLC/MS analyses were performed on a Waters ACQUITY UPLC H-Class system equipped with an autosampler and photodiode-array detector, coupled to a Bruker Impact II ultrahigh-resolution Qq-time-of-flight mass spectrometer controlled with Hystar 5.0 SR1 software.

Method 1 (Peptidoglycan).

LC separations were performed on a Waters ACQUITY UPLC HSS T3 column (100 Å, 1.8 μm, 2.1 mm × 150 mm) fitted with a VanGuard precolumn (100 Å, 1.8 μm, 2.1 mm × 5 mm) at 50 °C. A 40 min gradient was used: 5 min at 100% A, followed by a 35 min linear gradient to 20% B, where A = 0.1% formic acid in water and B = 0.1% formic acid in acetonitrile. During the first 4 min, the LC flow was diverted to waste. The electrospray source was operated in the positive-ion mode with an end-plate offset of −500 V, capillary voltage of 1800 V, and nitrogen as nebulizer gas (4 bar) and dry gas (7 L/min, 200 °C). Mass spectra were acquired over m/z 200–3000.

Method 2 (Intact Protein).

LC separations were carried out on a Waters ACQUITY UPLC Protein BEH SEC column (200 Å, 1.7 μm, 2.1 mm × 150 mm) with isocratic elution with 100 mM ammonium acetate as the mobile phase. The electrospray source was operated in the positive-ion mode with an end-plate offset of −500 V, capillary voltage of 2500 V, and nitrogen as both nebulizer gas (3 bar) and dry gas (6 L/min, 180 °C). Mass spectra were acquired over m/z 400–4500. Data were processed with Bruker Compass DataAnalysis 5.1, and multiple charged ions were deconvoluted using either the Bruker maximum-entropy algorithm or UniDec.⁴⁷

Scanning Electron Microscopy

Visualization of the morphological differences between wild-type P. aeruginosa MPAO1 and PA2854::Tn MPAO1 transposon-insertion mutant strains was made using scanning electron microscopy. Overnight growth of bacteria in liquid LB media was subcultured in liquid LB media, and 5 mL cultures were harvested by centrifugation (2000g for 10 min) during log-phase growth (OD₆₀₀ of 0.8). Harvested cells were then suspended in 10 mL of either phosphate-buffered saline pH 7.4 or water containing 10 mM ethylenediaminetetraacetic acid (EDTA) for 5 min at room temperature for either isotonic or hypotonic treatment. The cells were then pelleted and washed twice by brief resuspension in the same treatment solution and pelleting by centrifugation. The pelleted cells were resuspended in 100 μL of the treatment solution and fixed on a poly-l-lysine-coated glass coverslip using 2% glutaraldehyde and 1% OsO₄ in 0.1 M sodium cacodylate buffer pH 7.5. The samples were dehydrated using ethanol, fixed on aluminum stubs, and sputter-coated with a 3 nm iridium coat. The samples were imaged on a Magellan 400 XHR electron microscope.

Cloning, Overexpression, and Purification of PA2854 and OprI

Constructs of the PA2854 coding sequence from P. aeruginosa PA01 were designed, excluding the N-terminal signal peptide (residues 1–23). The construct used in this study was the full-length form (amino acids 24–323). The preparations of the clone were described.² The construct included a histidine tag handle MGSSHHHHHHSSGEN-LYFQGHM to enable purification by nickel affinity chromatography. Protein expression and purification followed established procedures.² The N-terminal His-tagged full-length construct (His-24–323) successfully yielded crystals suitable for structure determination. This protein was maintained in a final buffer of 50 mM Tris pH 8 and 150 mM NaCl, with the histidine tag retained during crystallization, as removal by TEV protease led to significant protein precipitation. For functional assays, a portion of the protein was processed with a 50:1 w:w ratio of recombinant protein to the TEV protease overnight at 4 °C, followed by removal of the His-tag using Ni-NTA resin to produce PA2854 lacking the N-terminal His-tag.

The cloning, expression, and protein purification for OprI (aa 25–83) have been described previously.³ Briefly, the oprI coding sequence corresponding to amino acids 25–83 was amplified by PCR from the P. aeruginosa PAO1 genome and inserted into pET41ST, which produced the fusion protein GST-6 His-tag-OprI (25–83) with a TEV cleavage site. Following purification of the fusion protein using Ni-NTA and glutathione resin (Genscript), the tag was removed by cleavage with the TEV protease. The tags could then be separated from untagged OprI using Ni-NTA.

Documentation of OprI Trimerization by Native-Mass Spectrometry

The methodology for nondenaturing mass spectrometry and buffer exchange has been described previously.^48,49 Briefly, OprI in 1 M ammonium acetate, pH 6.6 (5–20 μM), was analyzed on a Bruker MicrOTOF-Q II quadrupole time-of-flight hybrid mass spectrometer. The electrospray ionization source was operated in the positive-ion mode with the following parameters: end-plate offset voltage −500 V, capillary voltage 4200 V, and nitrogen as both a nebulizer (1.2 bar) and dry gas (3 L/min, 180 °C), funnel 1 RF 400 V, funnel 2 RF 500 V, hexapole RF 600 V, quadrupole ion energy 4 eV, collision energy 6 eV, collision cell RF 2400 V, ion transfer time 100 μs, and prepulse ion storage 25 μs. Mass spectra were acquired over the mass range m/z 500–10,000, and mutiply charge ions were deconvoluted using UniDec.⁴⁷

Thermal-Shift Assessment of PA2854 Binding to Its Substrates

Thermal-shift assays were conducted using nanodifferential scanning fluorimetry (nanoDSF) on a Tycho NT.6 instrument (NanoTemper). PA2854 (1.5 mg/mL) was incubated with OprI and PG substrates 6 or 7, and their combinations in 100 mM Tris pH 8.0, and 1 mM β-mercaptoethanol for 1 h on ice. Samples were then loaded onto Tycho NT.6 capillaries and heated from 35 to 95 °C. Fluorescence at 330 and 350 nm was monitored, and the melting temperature of the protein was taken as the inflection in fluorescence ratio (350 nm/330 nm). Control samples of PA2854 and OprI were also measured for reference. Data plotting and statistical analysis were done in GraphPad Prism.

Enzymes and Peptidoglycan Substrates

To prepare the muropeptide mixture of 6 and 8, trypsin-treated sacculus of E. coli K12 cells was digested with MltA in buffer containing 10 mM sodium acetate pH 5.0, 10 mM MgCl₂, and 0.05% Triton X-100. For preparing the PG mixture of 7 and 9, the trypsin-treated sacculus was digested with lysozyme in buffer containing 10 mM sodium acetate pH 6 and 0.05% Triton X-100. After digestion, the reaction mixtures were passed through a 10 kDa MWCO filter, and the filtrates were lyophilized, reconstituted in water, passed through a detergent removal column, and lyophilized again.

Reaction of PA2854 Using Synthetic Substrates and Meropenem

We attempted to reconstitute the PA2854 reaction using different buffer (Tris, phosphate-buffered saline, and acetate) at various pH values (pH 6–8). We also attempted the reaction at 4 °C, 25 °C, and 37 °C. We report here the conditions under which the reaction was successful. PA2854 and OprI in Tris buffer were buffer-exchanged into 50 mM HEPES, pH 7.5, using a Zeba desalting column (7 kDa MWCO), according to the manufacturer’s instructions. For PA2854–Meropenem reactions, PA2854 (20 μM) was incubated with Meropenem (4 mM) in 10 mM HEPES, pH 7.5, 100 mM NaCl, and 0.1 mM tris(2-carboxyethyl)phosphine (TCEP) at 4 °C for 72 h. Aliquots were withdrawn at 11 time points and analyzed by UHPLC/MS method 1; all reactions were performed in duplicate. For PA2854–OprI–peptidoglycan reactions, PA2854 (20 μM), peptidoglycan fragments (4–7, 6 + 8 mix, or 7 + 9 mix) at 2 mM, and OprI (50 μM) were incubated in 10 mM HEPES, pH 7.5, 100 mM NaCl, and 0.1 mM TCEP at 4 °C for up to 96 h. These reactions were also performed in duplicate and analyzed by the UPLC/MS method 2.

Crystallization of PA2854 and OprI

The purified His-tagged full-length PA2854 construct was concentrated to 6 mg/mL using a 10 kDa Amicon Ultra Centrifugal filter. High-throughput crystallization screening was conducted at 18 °C in 96-well plates, using commercial crystallization screens from Hampton Research, Qiagen, Molecular Dimensions, and Jena Bioscience, and set up with the Oryx8 Crystallization Robot. The best crystals of PA2854 were obtained with 100 mM sodium acetate pH 4.6 and 1.1 M diammonium tartrate, using sitting-drop vapor diffusion with 150 μL reservoir solution and drops composed of 1 μL of protein mixed with 1 μL of the precipitant. Additional optimization efforts—including the use of additives, detergents, and alternative methods such as hanging-drop and microbatch techniques at both 18 °C and 4 °C—did not lead to improved crystal quality. Crystals were cryoprotected by soaking in the reservoir solution supplemented with 15% glycerol prior to flash-cooling in liquid nitrogen.

OprI (amino acids 25–83) was concentrated up to 18.8 mg/mL using 3 kDa Amicon Ultra Centrifugal filters. The best crystal was obtained in the sitting-drop vapor diffusion method in a condition with 30% PEG 200, 0.1 M MES pH 6.5, and 0.1 M sodium acetate. The crystals grew in drops of 0.2 μL of protein and 0.2 μL of precipitant equilibrated against 65 μL of the reservoir solution at 18 °C. Crystals were directly flash-cooled in liquid nitrogen.

X-ray Data Collection and Structural Determination of PA2854 and OprI

X-ray diffraction data were collected for PA2854 at the XALOC beamline of the ALBA Synchrotron (Barcelona, Spain), using a wavelength of 0.979 Å and a Pilatus 6 M detector. The diffraction pattern exhibited pronounced anisotropy, reflected in varying diffraction quality along different reciprocal lattice directions. Data processing was performed using XDS,⁵⁰ followed by ellipsoidal truncation with the STARANISO server.⁵¹ The diffraction-limit surface was defined using a CC_1/2 > 0.3 threshold. The resulting resolution limits along the principal axes of the ellipsoid cutoff were: a* = 3.84 Å, b* = 3.84 Å, and c* = 2.44 Å, providing an overall resolution of 2.63 Å. Phase determination was achieved by molecular replacement, using an AlphaFold2-predicted model³⁷ as the search template. Model building and refinement were conducted manually in Coot, Phenix,⁵² and Refmac5.⁵³ PA2854 crystallized in space group P4₃2₁2, with one molecule per asymmetric unit and a solvent content of 72%. Unit-cell parameters and detailed data collection and refinement statistics are summarized in Supporting Table S1. Structural visualization was carried out with PyMOL^54,55 and ChimeraX.⁵⁶

Diffraction data were collected for OprI at the ID30-A3 beamline of the ESRF Synchrotron (Grenoble, France), using a wavelength of 0.8731 Å and a Eiger1 × 4 M detector. Data were reduced using XDS. CC_1/2 > 0.3 was used as a criterion to determine the resolution limit. OprI crystallized in the cubic space group P2₁3 with one molecule per asymmetric unit. Structure determination was achieved with the molecular replacement method using Alphafold2 model. Model building and refinement were conducted in Coot and Refmac5. Statistics are presented in Supporting Table S1. Figures were created using PyMOL and ChimeraX.

Computational Modeling of PA2854:OprI-Trimer Complex

The structure of the noncovalent PA2854:GlcNAc-MurNAc-tetrapeptide complex was generated using RoseTTAFold All-Atom.⁴¹ The predicted coordinates of GlcNAc-MurNAc-tetrapeptide were merged with the crystallographic coordinates of PA2854. The complex was prepared and energy-minimized using the protein preparation wizard⁵⁷ of Schrödinger Suite 2024–4 at pH 7.4. The terminal d-Ala was then deleted, and the thioester bond between m-DAP and Cys207 was manually constructed to give the covalent PA2854:Glc-NAc-MurNAc-tripeptide complex. This thioester complex was prepared using the system builder of Schrödinger Suite 2024–4 in an orthorhombic TIP3P water box with buffer distances of 10 Å. The system was neutralized, and NaCl was added to a concentration of 0.15 M. The system was then simulated in Desmond⁵⁸ for 20 ns in an NPT ensemble to allow for its relaxation. A representative cluster of the simulation was used for building the full PA2854:OprI-Trimer:GlcNAc-MurNAc-tripeptide complex.

In our crystallographic coordinates of OprI, a 7-amino acid stretch at the N-terminus (CSSHSKE) is not resolved. Modeler⁵⁹ was used to construct the missing N-terminal sequence on the trimeric structure, while preserving the helical structure of the template crystallographic coordinates of OprI. The full-length coordinates of trimeric OprI were then solvated and neutralized as described for PA2854 and simulated for 10 ns. A representative cluster of the simulation was used for building the full PA2854:OprI trimer:GlcNAc-MurNAc tripeptide complex.

Targeted Molecular Dynamics Simulation of PA2854

The transition of PA2854 from the latent crystallographic conformation to the active conformation shown in the AlphaFold model was studied by targeted molecular dynamics. These simulations were performed using Amber 24.0.^60,61 The structures of the latent and active conformations of PA2854 were treated as the initial and target conformations, respectively. Both structure files in PDB format were prepared using tLEaP using an FF19SB force field, neutralized, and solvated in a ~15 Å octahedral TIP3P water box. Each structure was then energy-minimized over three stages. In the first stage, only water and ions were allowed to move. In the second stage, the entire system was mobile, but a 0 kcal·mol⁻¹·Å⁻² harmonic restraint force constant was applied to the protein. Finally, in the third stage, the entire system was energy-minimized with no restraint force. Each system is then equilibrated over five stages. In the first stage, the system is heated to 300 K, and only solvent and water movement took place. Second, the system is equilibrated with no protein movement. In the third to final equilibration stages, the system is equilibrated in NPT with sequentially decreasing the harmonic restraint force constant from 50 to 20 to 5 kcal·mol⁻¹·Å⁻². These preparation steps yield the simulation-ready system.

Targeted molecular dynamics was completed over three consecutive simulations totaling 11 ns. In the first simulation, a mild harmonic restraint force of 0.25 kcal·mol⁻¹·Å⁻² was applied to the backbone of loops L_C, L_122–133, and L_249–263 for 5 ns. Next, the system was simulated for 3 ns with a moderate restraint force of 0.5 kcal·mol⁻¹·Å⁻² applied only to all heavy atoms of L_c. Finally, all heavy atoms of all three loops were once again subjected to a restraint of 0.5 kcal·mol⁻¹·Å⁻² for 3 ns of simulation time. All three simulations were combined using CPPTRAJ and visualized using PyMOL.^62-64 The full targeted molecular dynamics simulation is shown in Movie S1 made with PyMOL.

Complex Generation and Molecular Dynamics Simulations

To construct the full PA2854:OprI-trimer:GlcNAc-MurNAc-tripeptide, the crystallographic coordinates were used in conjunction with the modeled AlphaFold-Multimer structure of PA2854:OprI-trimer. The complex preserves the crystallographic details of the proteins, while mimicking their mode of interaction predicted by AlphaFold-Multimer. The reconstructed system (PA2854:OprI-trimer:GlcNAc-MurNAc-tripeptide) was solvated and ionized as previously described⁶⁵ and simulated for 20 ns in an NVT ensemble. A 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) membrane was generated using the Schrödinger Suite 2024–4 system builder and covalently linked to the N-terminal cysteine of OprI to model the natural anchoring of OprI to the outer membrane of P. aeruginosa. The full complex was solvated, neutralized, and simulated for 100 ns in the NVT ensemble. The resulting trajectory was analyzed using PyMOL^60,61 (Movie S2). The angle of nucleophilic attack and distance between the ε-amino nitrogen of N-terminal lysine and the thioester carbonyl were monitored over the duration of the simulation and plotted using GraphPad Prism.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.6c03160.

• Non-OprI-linked and OprI-linked muropeptide structures and fragment nomenclature (Figure S1); SDS-PAGE and mass spectrometry analyses of OprI and PA2854 (Figure S2); Reaction scheme and products of OprI-PG cross linking (Figure S3); Mass spectrometry analysis of PA2854 reaction with Meropenem (Figure S4); Electron-density and flexibility of PA2854 and OprI (Figure S5); Structural homologs of P. aeruginosa PA2854 (Figure S6); Conservation of the LD-transpeptidase motif (Figure S7); AlphaFold2 prediction for PA2854 (Figure S8); Structural comparison of OprI and Lpp (Figure S9); AlphaFold2-predicted models of PA2854 in complex with OprI (Figure S10); The surface electrostatic potential of the acceptor-binding site in PA2854, YkuD, Ldtfn217 and YcbB (Figure S11); Proposed model of outer membrane-PG anchoring in P. aeruginosa (Figure S12); Crystallographic data collection and refinement statistics for PA2854 and OprI (Table S1) (PDF)

• Targeted molecular-dynamics simulations of the activation of PA2854 (MP4)

• Molecular-dynamics simulations of the PA2854 catalytic complex (MP4)

ABBREVIATIONS

PG

peptidoglycan

GlcNAc-MurNAc

β-(1 → 4)-N-acetylglucosamine-N-acetylmuramic acid

m-DAP

meso-diaminopimelate

LC-MS

liquid chromatography–mass spectrometry

TIC

total-ion chromatogram

Tetra

N-acetylglucosamine-β-1,4-N-acetylmuramic acid with tetrapeptide

Tri

N-acetylglucosamine-β-1,4-N-acetylmuramic acid with tripeptide

MurNAc_r

reduced N-acetylmuramic acid

anhMurNAc

1,6-anhydro-N-acetylmuramic acid

EICs

extracted-ion chromatograms

CID

collision-induced dissociation

EDTA

ethylenediaminetetraacetic acid

K _D

equilibrium-dissociation constant

HEPES

(4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid)

MEM

meropenem

nanoDSF

nanodifferential scanning fluorimetry

ΔT_i

shift in protein melting temperature

TEV

tobacco-etch virus

Ldt

ld-transpeptidase

RMSD

root-mean-square deviation

POPE

1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine

SDS

sodium dodecyl sulfate