Mechanism of anchoring of OmpA protein to the cell wall peptidoglycan of the gram-negative bacterial outer membrane

Abstract

The outer membrane protein A (OmpA) plays important roles in anchoring of the outer membrane to the bacterial cell wall. The C-terminal periplasmic domain of OmpA (OmpA-like domain) associates with the peptidoglycan (PGN) layer noncovalently. However, there is a paucity of information on the structural aspects of the mechanism of PGN recognition by OmpA-like domains. To elucidate this molecular recognition process, we solved the high-resolution crystal structure of an OmpA-like domain from Acinetobacter baumannii bound to diaminopimelate (DAP), a unique bacterial amino acid from the PGN. The structure clearly illustrates that two absolutely conserved Asp271 and Arg286 residues are the key to the binding to DAP of PGN. Identification of DAP as the central anchoring site of PGN to OmpA is further supported by isothermal titration calorimetry and a pulldown assay with PGN. An NMR-based computational model for complexation between the PGN and OmpA emerged, and this model is validated by determining the crystal structure in complex with a synthetic PGN fragment. These structural data provide a detailed glimpse of how the anchoring of OmpA to the cell wall of gram-negative bacteria takes place in a DAP-dependent manner.—Park, J. S., Lee, W. C., Yeo, K. J., Ryu, K.-S., Kumarasiri, M., Hesek, D., Lee, M., Mobashery, S., Song, J. H., Lim, S. I., Lee, J. C., Cheong, C., Jeon, Y. H., Kim, H.-Y. Mechanism of anchoring of OmpA protein to the cell wall peptidoglycan of the gram-negative bacterial outer membrane.

Acinetobacter baumannii is an important multidrug-resistant nosocomial pathogen that causes pneumonia, meningitis, bacteremia, and wound and urinary tract infections, among other complications, in immunocompromised patients (1). Studying the structure and function of A. baumannii outer membrane protein A (AbOmpA) is of interest because it is the most abundant surface protein, and it plays a role in the permeability of antibiotics and small molecules and is also a key player in bacterial pathogenesis, inducing host cell death (2–4). An AbOmpA-deficient mutant showed a significant increase of the rate of host cell viability and the disappearance of the outer membrane of A. baumannii (5). Gram-negative bacteria secrete outer membrane vesicles (OMVs) as vehicles for the transport of effector molecules, such as virulence factors and immune modulators, into host cells as part of their pathogenesis (6, 7). AbOmpA is one of the virulence-associated proteins in the OMVs and plays an important role in the formation of OMVs (8). It has been suggested that the outer membrane protein A (OmpA)-like domain of OmpA interacts with peptidoglycan (PGN) and that it may control OMV production and the stability of gram-negative bacteria (9); however, details of its structure are not clear.

PGN is the major constituent of the bacterial cell wall and provides structural strength and controls cell shape, and its integrity is critical for bacterial survival (10, 11). PGN is a polymer made up of repeating disaccharide units of N-acetylglucosoamine (NAG) and N-acetylmuramic acid (NAM). A pentapeptide stem is appended to the NAM moiety, which is the site of cross-linking with neighboring PGN strands (12–14). The typical pentapeptide in most gram-negative bacteria is l-Ala-γ-d-Glu-meso-DAP-d-Ala-d-Ala (DAP represents diaminopimelate, a unique bacterial amino acid). DAP is replaced by l-Lys in many gram-positive bacteria (15). These two forms are common, yet other organisms have evolved subtle variations on these two themes.

In gram-negative bacteria, the cell wall is located between the inner membrane and the outer membrane. The mechanism by which the cell wall interacts with these membranes is largely unknown, but the OmpA family of proteins has been proposed to interact noncovalently with PGN (16). The interaction of this protein with the cell wall is believed to provide stability to the supramolecular assembly and result in cellular integrity. The OmpA family encompasses 3 distinctive groups of proteins (Supplemental Fig. S1): outer membrane proteins (OmpA, NmRmpM, CadF, OprF, and others), PGN-associated lipoproteins (Pal, OprL, and others), and bacterial flagellar motor proteins (MotB, MotY, SciZ, and others). Outer membrane proteins and PGN-associated lipoproteins serve as anchors between the outer membrane and PGN structure (17–20). Bacterial flagellar motor proteins link PGN with the inner membrane to immobilize the stator ring of the bacterial flagellar motor (21, 22). OmpA-like proteins exist in both gram-negative and gram-positive bacteria that contain DAP-type PGN.

Outer membrane proteins in gram-negative bacteria serve a number of functions, including uptake of nutrients and bacterial pathogenesis, and are attractive as potential vaccine candidates (16, 23, 24). The OmpA protein from Escherichia coli has been studied extensively and is composed of an N-terminal membrane-embedded β-barrel domain and a C-terminal OmpA-like domain, which has been proposed to interact specifically with PGN (25). Even though the structures of various OmpA-like domains have been determined (20, 22, 26), there is a paucity of information on the structural aspects of the mechanism of PGN recognition by the OmpA-like domain because of the lack of high-resolution X-ray crystallographic structures in complex with PGN components. Herein, we report our work to elucidate the molecular recognition process in A. baumannii. In the present study, AbOmpA-periplasmic domain (AbOmpA-PD) (residues 221–339) was cloned into E. coli, expressed, and purified. We determined the high-resolution crystal structure of AbOmpA-PD bound to DAP. Binding studies in combination with investigations of NMR chemical shift perturbations with several synthetic PGN fragments allowed the mapping of the binding surface and led to a computational model for the binding of PGN with AbOmpA-PD. The crystal structure of AbOmpA-PD in complex with a synthetic PGN fragment validated the NMR-based computational model. Our results provide structural information on how the PGN is anchored to the outer membrane, and these findings should have a profound impact on determining the structural details of the assembly of the bacterial envelope.

MATERIALS AND METHODS

Cloning, expression, and protein purification

The AbOmA-PD gene was cloned from the full-length AbOmpA gene into the pET28a vector and overexpressed in E. coli as a His₆-tagged fused protein. Cells were harvested and suspended in a lysis buffer (20 mM Tris-HCl and 1 mM PMSF, pH 8.0) and then were sonicated on ice and ultracentrifuged at 200,000 g for 1 h. The supernatant was denatured for 2 h at room temperature by shaking in a solubilization buffer (6 M guanidine hydrochloride, 20 mM Tris-HCl, 500 mM NaCl, 5 mM imidazole, and 1 mM 2-mercaptoethanol, pH 8.0). After centrifugation of the denatured mixture, the soluble fraction including the His₆-tagged recombinant proteins was collected and loaded on a 5-ml HisTrap column (GE Healthcare, Little Chalfont, UK) that had been preequilibrated with solubilization buffer. Unbound proteins were washed off with refolding buffer (6 M urea, 20 mM Tris-HCl, 500 mM NaCl, 5 mM imidazole, and 1 mM 2-mercaptoethanol, pH 8.0), and the bound protein was refolded on a column using a linear gradient (300 ml in 0.5 ml/min) into wash buffer (20 mM Tris-HCl, 500 mM NaCl, 10 mM imidazole, and 1 mM 2-mercaptoethanol, pH 8.0). The refolded proteins were eluted using an elution buffer (20 mM Tris-HCl, 500 mM NaCl, 500 mM imidazole, and 1 mM 2-mercaptoethanol, pH 8.0). All of the refolding and elution steps were performed with ÄKTAprime (GE Healthcare). The selenomethionyl protein used for the single-wavelength anomalous diffraction (SAD) experiments was prepared as described above.

PGN derivatives

PGN analogs were synthesized [NAM-tripeptide (compound 4) and NAM-pentapeptide (compound 5)] as described in ref. 27. DAP was obtained from Sigma-Aldrich (St. Louis, MO, USA). Tripeptide DAP (compound 2) and tripeptide Lys (compound 3) were purchased from InvivoGen (San Diego, CA, USA).

Crystallization and structure determination

AbOmpA-PD crystals were grown at 293 K using the hanging-drop vapor-diffusion method. The crystals were obtained by mixing the protein solution (20 mg/ml) with the same volume of the well buffer containing 15–20% PEG 3350, 0.1 M HEPES (pH 8.0), and 0.2 M ammonium sulfate. The crystals belonged to space group P2₁ and contained 8 molecules in the asymmetric unit. To obtain crystals of the OmpA-compound 5 complex, OmpA was incubated on ice with a 2-fold molar excess of compound 5 before crystallization in a buffer containing 62.5% (v/v) 2-methyl-2,4-pentanediol and 0.1 M HEPES (pH 7.5). The native, complex, and SAD data sets were collected using a synchrotron radiation source at beamlines 6C1 and 4A in the Pohang Accelerator Laboratory (Pohang, Korea) and beamline 17A in the Photon Factory (Tsukuba, Japan). Data were processed using the program HKL2000 (28). The structure of AbOmpA-PD was determined by the SAD method using data collected with selenomethionyl protein crystals and the program Solve (29). The phase was further improved by the program Resolve (30), and the initial model was built automatically by the program ARP/wARP (31). The data collection statistics and refinement statistics are listed in Table 1. The model was completed using iterative cycles of model building with Coot (32) and refinement with CNS (33, 34) and Refmac5 (35). The crystal structures of the OmpA-DAP and OmpA-compound 5 complexes were determined using molecular replacement methods and the program MolRep (36).

Table 1.

Data collection and refinement statistics

Statistic	Selenomethionine	Native	DAP	Compound 5
Data collection
Space group	P21	P21	P21	P21
Cell dimensions
a, b, c (Å)	58.3, 98.7, 98.7	58.1, 98.8, 98.3	57.9, 98.2, 98.4	65.1, 162.6, 66.2
α, β, γ (deg)	90.0, 105.9, 90.0	90.0, 105.8, 90.0	90.0, 105.5, 90.0	90.0, 112.7, 90.0
Wavelength (Å)	0.97946	1.0000	1.0000	0.98
Resolution (Å)	50–2.1	94.6–1.59	94.2–1.8	81.4–2.0
Rsyma	0.077 (0.220)	0.051 (0.373)	0.082 (0.437)	0.102 (0.349)
I/σI	16.4 (3.2)	20.4 (2.4)	18.7 (2.1)	9.1 (3.96)
Completeness (%)	95.7 (92.4)	99.7 (99.7)	99.6 (97.0)	92.2 (94.5)
Redundancy	6.6	2.6	3.7	2.7
Refinement
Reflections		133,854	93,970	78,621
Rwork/Rfreeb		0.190/0.230	0.185/0.224	0.220/0.257
Atoms (n)
Protein		7729	7738	7950
Water		1106	834	885
B factor (Å2)
Protein		15.23	16.68	20.84
Ligand		12.32	25.39	17.40
Water		26.90	29.96	28.97
Root mean square deviation
Bond lengths (Å)		0.031	0.026	0.005
Bond angles (deg)		2.524	2.202	1.175

Values in parentheses correspond to the highest-resolution shell.

^aR_sym = Σ |I − 〈I〉|/Σ I.

^bR factor = Σ ‖F_obs(hkl)| − |F_calc(hkl)‖/Σ |F_obs(hkl)|.

Isothermal titration calorimetry (ITC)

ITC measurements were performed at 298 K on a VP-ITC system (MicroCal, Piscataway, NJ, USA) with a reference power of 10 cal/s and a stirring speed of 307 rpm. Protein samples were dialyzed against ITC buffer containing 20 mM Bis-Tris (pH 6.8) and 50 mM NaCl for 12 h at 277 K. Ligand samples were dissolved in the dialysis buffer. For the ITC experiment, m-DAP (d,l-DAP, Sigma) was used instead of the DAP mixture. Glycine and glycinamide were purchased from Sigma-Aldrich. All sample solutions were centrifuged at 18,000 g for 20 min and thoroughly degassed for more than 20 min using moderate stirring under vacuum. The 1.4-ml sample cell was filled with 10–30 μM protein, and the 300-μl injection syringe was filled with 200–600 μM ligand. Each titration typically involved 30 injections with 8 μl, respectively, at 16-s durations and 240-s intervals. The data fitting and analysis were performed using the Origin 7.0 software package provided by MicroCal.

PGN-binding assay

The assay was performed as reported previously (37). In brief, 10 μg of purified AbOmpA-PD and 125 μg of commercially available PGNs from Staphylococcus aureus (InvivoGen), Bacillus subtilis (Fluka, Buchs, Switzerland), or E. coli (InvivoGen) were dissolved in 100 μl of binding buffer containing 20 mM Tris-HCl (pH 7.8) and 300 mM NaCl and incubated on a shaking platform for 24 h at 273 K. Bound protein, which was retained in the PGN pellet after the incubation mixture was centrifuged at 200,000 g for 1 h, was washed 3 times with 500 μl of binding buffer, centrifuged for 15 min at 15,000 rpm and finally dissolved in 20 μl of SDS sample buffer for electrophoresis. The PGN-bound proteins were visualized by Coomassie Brilliant Blue R-250 (Sigma-Aldrich) staining.

NMR experiments

For the NMR experiments, ¹⁵N-labeled or ¹³C/¹⁵N-labeled AbOmpA-PD protein was overexpressed by adding 1 mM isopropyl-β-d-1-thiogalactopyranoside for ∼24 h at 291 K in M9 minimal medium enriched with ¹⁵NH₄Cl as the sole nitrogen source (99% ¹⁵N; Cambridge Isotope Laboratories, Andover, MA, USA) using either ¹³C-labeled (99%, U-¹³C₆; Cambridge Isotope Laboratories) or unlabeled glucose (Sigma-Aldrich) as the sole carbon source.

All NMR spectra were recorded using an Avance 900-MHz NMR spectrometer equipped with a triple-resonance cryoprobe (Bruker, Billerica, MA, USA). The ¹H-¹⁵N heteronuclear single quantum correlation (HSQC) spectra of 0.5 mM ¹⁵N-labeled AbOmpA-PD combined with 1 mM ligand (DAP and compounds 2, 4, and 5) were acquired at 298 K in a buffer containing 20 mM Bis-Tris (pH 6.6) and 100 mM NaCl. The backbone assignments of ¹³C/¹⁵N-labeled AbOmpA-PD with DAP and compound 5 were made by 3-dimensional heteronuclear correlation experiments: HNCO, HNCA, HNCACB, and CBCA(CO)NH. Data were processed with NMRPipe (38) and analyzed with the Sparky program (39). The magnitude of the ¹H-¹⁵N chemical shift differences (Δδ, ppm) were calculated using the equation Δδ = {(δH²)+0.2×(δN²)}^1/2, where δH and δN are changes to the proton (¹H) and nitrogen (¹⁵N) chemical shifts, respectively (40).

Computational methods

The X-ray structure of OmpA-PD complexed with DAP provided the initial cartesian coordinates. The structure of compound 5 was constructed using the Sybyl molecular modeling program (Tripos Associates Inc., St. Louis, MO, USA) and docked flexibly to the OmpA-PD structure, for which the crystallographic coordinates for DAP were removed. The poses of the ligand in complex with the protein were scored using the Goldscore scoring function of the program Gold 4.2 (41, 42). In addition, to reproduce the binding mode of DAP seen in the crystal structure, the side-chain carboxylate of DAP in the ligand was constrained to the vicinity of the two R286 Nη atoms. The highest-scored pose on the OmpA X-ray structure was entirely consistent with the NMR signal Δδ values, and this pose was selected for further studies. The experimental chemical Δδ values for the complexes were assessed for 5 different sets: DAP vs. compound 2, DAP vs. compound 4, DAP vs. compound 5, compound 2 vs. compound 4, and compound 2 vs. compound 5. To assess the distribution of the locations of NMR differences on the tertiary structure of OmpA-PD in the direction of greater complexity seen in the ligands, all residues exhibiting Δδ ≥ 0.05 were mapped and color-coded onto the X-ray structure using Sybyl. Finally, the selected pose for the ligand was incorporated into the backbone of NAG- NAM(pentapeptide)-NAG-NAM(pentapeptide) for which the solution NMR structure was determined (14). The first peptide in the ligand model for docking and scoring interacts with the protein (referred to as the bound peptide hereafter), whereas the second peptide points in the direction of the milieu. The model was then optimized while keeping the protein atom coordinates restrained using Sybyl and the Tripos force field parameters with Gasteiger-Hückel atomic charges. These optimized atom positions served as the initial coordinates for molecular dynamics (MD) simulations.

MD simulations

Hydrogen atoms were added to the optimized OmpA-PD complex using Sybyl, and atomic charges for the tetrasaccharide with the two peptide stems were determined using a 2-stage RESP procedure (43). Amber ff99 and GAFF provided force field parameters for the protein and the ligand, respectively. The complex was immersed in a truncated octahedral periodic box of TIP3P (44) water molecules and neutralized. The solvent was subjected to 50,000 steps of energy minimization followed by 100 ps of equilibration while the protein-ligand complex was restrained. The restraints were then removed gradually, and the system was heated to 300 K over a period of 100 ps at a constant pressure of 1 atm. Finally, the Sander module in Amber 9.0 (45) was used to perform MD simulations, and data were collected every 0.5 ps over a 10-ns trajectory. The averaged structure of the complex was computed on the basis of snapshots of the final 5 ns, and it was energy minimized using the conjugated-gradient method for 80,000 steps.

RESULTS

Overall structure and DAP binding mode of AbOmpA-PD

The crystal structure of AbOmpA-PD (residues 221–339, Fig. 1A) was determined using the SAD of selenomethionine-labeled protein crystals, for which the native structure was modeled and refined to a resolution of 1.6 Å. We also determined the crystal structure of AbOmpA-PD complexed with DAP at the resolution of 1.8 Å and complexed with a synthetic PGN fragment at 3.0 Å. The data processing and refinement statistics are summarized in Table 1.

Figure 1.

Sequence alignment of OmpA-like domains and the crystal structure of AbOmpA-PD. A) Multiple amino acid sequence alignment of AbOmpA homologs. Secondary structure elements from the AbOmpA structure are indicated above the alignment. Residue numbers are for AbOmpA, and red blocks denote regions of sequence identity across all homologs. Partially conserved residues are boxed. Residues important for DAP binding by AbOmpA are highlighted with stars below the sequence alignment. PaOprF, Pseudomonas aeruginosa OprF; CjCadF, Campylobacter jejuni CadF; EcPal, E. coli Pal; HiPal, H. influenzae; PaOprL, P. aeruginosa Pal; HpMotB, H. pylori MotB; VaMotY, Vibrio alginolyticus MotY; EcSciZ, E. coli SciZ. B) Ribbon depiction of AbOmpA-PD in complex with DAP. Model was rendered on the basis of its secondary structure, in which α helices and β strands are shown in red and yellow, respectively. Structure of bound DAP is drawn as a space-filled model, and atoms are colored based on atom type (nitrogen, blue; oxygen, red; carbon, gray). C) Surface electrostatic potential depiction of AbOmpA (red, negative; blue, positive). Residues responsible for the basic patches that would interact with the PGN on the surface are indicated by the dotted area.

The overall structure of AbOmpA-PD belongs to the OmpA-like domain-fold family, which consists of a 4-stranded β sheet and 3 α helices that are arranged in the topological order β/α/β/α/β/α/β (Fig. 1B). In the native structure, 1 glycine molecule was observed in a small cavity composed of helices α2 and α3 and the 3 loops β1α1, β2α2, and α3β4. In the DAP complex, the DAP molecule was embedded in the same cavity, with a carboxylate of DAP engaging the guanidinium Nη1 and Nη2 of Arg286 in a bidentate salt bridge with 2.8 and 3.0 Å separation, respectively. This carboxylate also forms hydrogen bonds to the main-chain amide protons of Asn237 (2.8 Å) and Asp271 (3.2 Å). In turn, Oδ of Asp271 is hydrogen bonded to Νζ of DAP (2.7 Å). In addition, Oδ1 of Asn279 and the backbone carbonyl group of Thr273 also bind to the Νζ of DAP, with distances of 2.8 and 2.9 Å, respectively (Fig. 2A).

Figure 2.

Stereo views of the AbOmpA-PD complex with DAP and superposed OmpA-like domains. A) Stereo representation of the closeup of the interaction between DAP (represented by its electron density in the omit map contoured at 2σ and represented in the model as the yellow capped stick within) and OmpA. Important residues within this binding pocket are rendered as capped sticks. Hydrogen bonds are indicated as dashed black lines. B) OmpA-like domains of AbOmpA, NmRmpM (PDB code 1R1M), EcPal (PDB code 1OAP), and HpMotB (PDB code 3CYQ) are superposed in the stereo ribbon depiction and shown in green, yellow, orange, and cyan, respectively. Strictly conserved Arg286 and Asp271 within this space are rendered as capped sticks.

Asp271 and Arg286 of AbOmpA-PD are important among the residues in contact with DAP, because they are completely conserved among the OmpA-like proteins (Fig. 1A). When we compared the 3-dimensional structures of these residues with the corresponding residues of NmRmpM [Protein Data Bank (PDB) code 1R1M], EcPal (PDB code 1OAP), and HpMotB (PDB code 3CYQ), they superimposed well (Fig. 2B), although the structures of NmRmpM, EcPal, and HpMotB do not contain DAP in the pocket. Hence, we suppose that the strictly conserved Asp271 and Arg286 residues are important for DAP binding to AbOmpA-PD. This assertion is supported by the site-directed mutagenesis substitutions of Asp271 and Arg286 by alanine in the AbOmpA-PD protein. As expected, the purified Asp271Ala and Arg286Ala mutants lost their ability to bind to DAP, as measured using ITC (Supplemental Fig. S2; compare Fig. 4B).

Figure 4.

PGN pulldown and ITC binding assays of meso-DAP with AbOmpA-PD. A) PGN pulldown results. B–D) ITC results with meso-DAP (K_d 2.2 μM; B), compound 2 (K_d 3.2 μM; C), and compound 3 (D).

We used a commercially available DAP mixture, composed of the 3 isomers, d,d-DAP, d,l-DAP, and l,l-DAP, for crystallization (Fig. 3). In the omit map, the amino acid group at C-6 that interacted with Arg286 was identified to be of the d form. However, in our structure, the amino group at C-2 of DAP did not interact with AbOmpA-PD, and there is no steric hindrance of the amino group at this position within the protein. Thus, it is not clear which stereoisomer—d,d-DAP or d,l-DAP—is bound in the complex structure. Because m-DAP is found in PGN of Acinetobacter sp. (46), the complex structure was modeled using d,l-DAP, which is compatible with the electron density in the omit map.

Figure 3.

Chemical structures of compounds used in this study. Compound 1, DAP; compound 2, tripeptide DAP; compound 3, tripeptide Lys; compound 4, NAM-tripeptide; compound 5, NAM-pentapeptide.

A significant cluster of positive charges is present near the DAP-binding cavity, and the adjacent surface shows mainly negative charges. The positively charged electrostatic potential is attributed to the basic residues Lys238 from loop β1α1; Arg276, Lys277, and Arg281 from α2; Lys320, Lys322, and Arg325 from α3; and Arg329 from loop α3β4 (Fig. 1C, dotted area). These positively charged residues comprise a surface for the interaction between AbOmpA-PD and PGN.

AbOmpA-PD binds specifically to DAP-type PGN

To investigate DAP-specific binding, we analyzed AbOmpA-PD binding to commercially available DAP- or lysine-type PGN using a pulldown assay and ITC. In PGN pulldown assays, AbOmpA-PD was found in the pellets of DAP-type PGN from both the gram-negative E. coli and the gram-positive B. subtilis, although the amounts of AbOmpA-PD pulled down were different for the two species, probably because of the different cell wall structures, such as amidation at the m-DAP of PGN in B. subtilis (15). In an experiment to assess the effect of amidation, glycinamide, an amidated form of glycine, failed to bind to AbOmpA-PD (Supplemental Fig. S3 and Supplemental Table S1). However, AbOmpA-PD was not detected in the pellets of lysine-type PGN from the gram-positive S. aureus (Fig. 4A). ITC assays revealed that wild-type AbOmpA-PD shows binding affinity to m-DAP (approximate K_d was 2.2 μM; Table 2) similar to that to compound 2 (Fig. 3; approximate K_d for the mixture of stereoisomers was 3.2 μM; Table 2), but did not bind to compound 3 (Fig. 4B–D and Table 2). These results indicate that AbOmpA-PD binds specifically to DAP-type PGN.

Table 2.

Thermodynamic parameters for the binding of AbOmpA-PD to PGN fragment from ITC experiment

Parameter	m-DAP	Compound 2	Compound 3
N	0.84 ± 0.01	1.59 ± 0.01	ND
K (M−1)	(45.8 ± 1.3) × 104	(31.7 ± 2.1) × 104	ND
ΔH (kcal/mol)	−61.9 ± 0.6	−31.6 ± 0.4	ND
ΔS (cal/mol/K)	−182	−80.6	ND
TΔS (kcal/mol)a	−54.2	−24.0	ND
ΔG (kcal/mol)b	−7.7	−7.5	ND
Kd (M)	(2.2 ± 0.1) × 10−6	(3.2 ± 0.2) × 10−6	ND

ND, not detected.

^aT = 298 K.

^bΔG = ΔH − TΔS.

Mapping of the PGN-binding site of AbOmpA-PD

To identify residues involved in interactions of AbOmpA-PD with PGN, 2-dimensional ¹H-¹⁵N HSQC spectra of the ¹⁵N-labeled AbOmpA-PD-DAP complex were compared with those of complexes with compounds 2, 4, or 5 (Supplemental Fig. S4). Compounds 4 and 5 were synthesized as meso-forms of DAP, retaining the d-stereo configuration at the C-6 position and the l-stereo configuration at the C-2 position (Fig. 3 and ref. 27). Binding of these pieces of PGN to the surface of AbOmpA-PD would cause chemical shift perturbations that provide information relating to the sites of interaction. For this analysis, the backbone ¹H, ¹⁵N, and ¹³C resonances of AbOmpA-PD were assigned using ¹⁵N/¹³C-labeled protein. We were able to assign resonances for 111 of 119 aa. The exceptions were the prolines and N-terminal residues. Details of the NMR assignments of the protein sequence will be published elsewhere.

Figure 5A gives the perturbations in chemical Δδ between the resonances of DAP-bound and PGN fragment (compounds 2, 4, and 5)-bound ¹⁵N-AbOmpA-PD. The colors magenta and orange on the structure shown in Fig. 5A represent large (Δδ>0.1 ppm) and moderate (0.1>Δδ>0.05 ppm) changes, respectively. Compound 2 perturbed the chemical shifts of residues in helix α2 and loops β1α1 and β2α2, indicating that the portion of the peptide composed of l-Ala-γ-d-Glu interacts with these sites (Fig. 5A, left panel). Compound 4 showed a similar effect, but also affected the resonances of Arg281, Leu282, and Ser289, all on the α2 helix. The additional effects are attributable to interactions with the NAM moiety (Fig. 5A, middle panel). Finally, the binding of compound 5 affected the resonances of Glu323 and Arg325 (located on helix α3), and the interaction with the middle portions of helix α2 became weaker (Fig. 5A, right panel).

Figure 5.

Chemical shift perturbations of AbOmpA-PD with various PGN derivatives and the computational model of the AbOmpA-PD complex with the PGN ligand. A) Differences in the ¹H-¹⁵N chemical shift (Δδ={δH²+0.2×(δN²)}^1/2) of DAP-bound AbOmpA-PD from those of the protein bound to compound 2 (left panel), compound 4 (middle panel), and compound 5 (right panel). Colors magenta and orange represent large (Δδ>0.1 ppm) and moderate (0.1>Δδ>0.05 ppm) changes in the chemical shifts, respectively. B, C) Stereo images of the computational model of PGN binding to AbOmpA-PD, represented using a ribbon model (B) and a solvent-accessible Connolly surface (C). PGN ligand is rendered as capped sticks; peptides are in blue; tetrasaccharide backbone is orange. Cavity that accommodates the side chain of DAP is at 6 o'clock in relation to the images in B and C.

These chemical shift perturbation results proved useful in a computational exercise designed to arrive at the structure for the complex of a PGN fragment with AbOmpA-PD. We used the NMR structure of the synthetic sample NAG-NAM(pentapeptide)-NAG-NAM(pentapeptide) for this effort (14). The previously determined structure shows the saccharide backbone as a right-handed helix, but beyond l-Ala the peptide is mobile and does not conform to a given structure. Figure 5B depicts the orientation of the ligand in the averaged and energy-minimized complex (based on MD simulation). This binding mode is consistent with all of the NMR chemical shift perturbation data that we measured and allows for fitting of the side chain of DAP within the cavity observed in the X-ray complex. Furthermore, the extension of the d-Ala-d-Ala terminus of the peptide to helix α3 also properly accounts for the perturbations observed at that binding site extreme. The PGN tetrasaccharide backbone aligns itself along helix α2 and the loop β1α1. This orientation suggests that the sugar backbone of PGN runs nearly parallel to the β-sheets of AbOmpA-PD. The same perspective for the solvent-accessible Connolly surface of AbOmpA-PD is illustrated in Fig. 5C and reveals the binding surface for the PGN on the protein.

The terminal carboxylate group of DAP in the bound peptide retains very similar hydrogen bonding interactions as in the X-ray crystal structure with the Arg286 Nη1 and Nη2; the N–O distances of 2.8 and 3.0 Å in the DAP-bound AbOmpA-PD Å become 2.8 and 2.9 Å, respectively, in the model. DAP Nζ interacts through hydrogen bonds with Asp271 Oδ (2.8 Å), Asn279 Oδ1 (2.8 Å), and Thr273 O (2.8 Å), and the respective distances in the X-ray crystal structure are 2.7, 2.8, and 2.9 Å. An additional finding that builds confidence in the model is that the terminal d-Ala of the peptide stem is nicely positioned to be electrostatically drawn to Arg325 as a terminal anchor. During MD simulation and energy minimization, the terminal ligand d-Ala carboxylate made two hydrogen-bonding contacts with Arg325 Nε and Nη2 (2.5 and 2.6 Å), a reasonably common motif for the recognition of carboxylates by the arginine guanidinium moiety.

With the NMR-based computational structure in hand, our continued attempts to obtain a single crystal of AbOmpA in complex with compound 5 succeeded. The structure of the complex was solved at 2.0 Å resolution. In the omit map of the structure, we observed clear electron density for the peptide portion of compound 5 (Fig. 6). The orientation of the peptide backbone was identified on the basis of the electron density that extends from γ-d-glutamate. No density was observed for the monosaccharide moiety, probably because of disorder. The overall structure of AbOmpA-PD complexed with compound 5 is in good agreement with the NMR model structure.

Figure 6.

Omit map (contoured at 2σ) of AbOmpA-PD (green) complexed with compound 5 (yellow).

DISCUSSION

Whereas the transmembrane domain of OmpA acts as a virulence factor, its periplasmic C-terminal domain associates with PGN and thus maintains the integrity of the outer membrane in concert with the protein Pal (47). AbOmpA is a major surface protein, and an AbOmpA-defective mutant shows significantly attenuated virulence and diminished integrity of the outer membrane in A. baumannii (5). In addition, the secretion of OMVs in A. baumannii (composed of PGN and outer membrane and periplasmic proteins) is an important mechanism of pathogenesis (8). Thus, the binding of AbOmpA-PD with PGN might also contribute to OMV stability. Hence, knowing how OmpA and the PGN interact may have implications for understanding the supramolecular assembly of the bacterial envelope and the mechanisms of pathogenesis.

Since the crystal structure of NmRmpM was first reported, a number of crystal structures of other OmpA-like proteins have been published (20, 22, 26) including those of E. coli Pal (EcPal, 29% amino acid sequence identity) and the C-terminal domain of Helicobacter pylori MotB (HpMotB, 33% identity) (Fig. 1A). The solution structure of Haemophilus influenzae Pal (HiPal, 31% identity) revealed that the OmpA-like domain of HiPal binds to a PGN peptide motif that specifically interacts with m-DAP (20). However, a detailed understanding of the mode of recognition of DAP did not emerge from these studies.

In the present study, we determined the crystal structure of AbOmpA-PD bound to DAP at 1.8 Å resolution. The interaction between AbOmpA-PD and DAP centers on Arg286 and Asp271, two highly conserved residues among OmpA-like proteins, which bind carboxylic and amine groups of DAP, respectively, and tightly anchor the d-amino acid moiety of DAP. Indeed, these two residues were found to be important in E. coli MotB function because the mutants of Asp197 and Arg217, corresponding to Asp271 and Arg286 of AbOmpA, respectively, were nonfunctional in vivo (48). In addition, we have provided evidence that AbOmpA-PD specifically binds to the DAP-type PGN using ITC and PGN pulldown assays. Using NMR chemical shift perturbations and computation, we propose a model of how AbOmpA-PD (an integral outer-membrane protein) anchors itself onto the PGN components of the cell wall. This model is consistent with the initial attraction of the negatively charged peptide of the PGN to the positively charged patch on the surface of AbOmpA-PD. This electrostatic attraction is followed by a highly engineered cavity that selectively accommodates the DAP moiety within the PGN to the exclusion of PGN variants that might have a Lys at this position. The structure of this complex provides a detailed glimpse of how anchoring of the outer membrane to the PGN favors DAP-type PGN over Lys-type PGN.

The carboxylate of m-DAP from B. subtilis PGN is known to be largely modified by amidation (15, 49), which might reduce the binding affinity of OmpA to m-DAP because of the diminished strength of the interaction between Arg286 guanidinium and m-DAP carboxamide. Because amidated m-DAP was not available commercially, an ITC binding assay of glycinamide and AbOmpA-PD was performed instead to test the effect of the amidation on the affinity of this interaction (Supplemental Fig. S3). Whereas glycine binds AbOmpA-PD with a K_d of 0.59 μM, glycinamide shows no detectable binding, confirming the decreased affinity by amidation (Supplemental Table S1). A similar result can be expected for amidated m-DAP, which requires the glycine moiety for the binding. In the pulldown assay, AbOmpA-PD binds B. subtilis PGN in a lesser amount than that of E. coli PGN, which might reflect the effect of amidation. It is noteworthy that the carboxylate form of m-DAP is still present in B. subtilis PGN despite amidation, probably in the primordial cell wall or loss of amidation during PGN preparation (49) and might have been detected in the pulldown assay.

Proper maturation of OMVs depends on the function of multiple outer membrane (OM) proteins. OM-PGN, which links proteins such as OmpA, has been implicated in the initiation of OMV formation by promoting the bulging of OM (9). Our structure depicts the detailed interactions that take place at the OM-PGN junction and suggests that the interaction between the OmpA-like domain and PGN can be perturbed by displacing DAP, a key component of PGN. This knowledge can be used to address infections from gram-negative pathogens, including A. baumannii, which pose a greater threat to the general public health (50).