Crystal Structures of a Quorum-Quenching Antibody

Summary

A large number of Gram-negative bacteria employ N-acyl homoserine lactones (AHLs) as signaling molecules in quorum sensing, which is a population density-dependent mechanism to coordinate gene expression. Antibody RS2-1G9 was elicited against a lactam mimetic of the N-acyl homoserine lactone and represents the only reported monoclonal antibody that recognizes the naturally-occuring N-acyl homoserine lactone with high affinity. Due to its high cross-reactivity, RS2-1G9 showed remarkable inhibition of quorum sensing signaling in Pseudomonas aeruginosa, a common opportunistic pathogen in humans. The crystal structure of Fab RS2-1G9 in complex with a lactam analog revealed complete encapsulation of the polar lactam moiety in the antibody combining site. This mode of recognition provides an elegant immunological solution for tight binding to an aliphatic, lipid-like ligand with a small head group lacking typical haptenic features, such as aromaticity or charge, which are often incorporated into hapten design to generate high-affinity antibodies. The ability of RS2-1G9 to discriminate between closely-related AHLs is conferred by six hydrogen bonds to the ligand. Conversely, cross-reactivity of RS2-1G9 towards the lactone is likely to originate from conservation of these hydrogen bonds as well as an additional hydrogen bond to the oxygen of the lactone ring. A short and narrow tunnel exiting at the protein surface harbors a portion of the acyl chain and would not allow for entry of the head group. The crystal structure of the antibody without its cognate lactam or lactone ligands revealed a considerably altered antibody combining site with a closed binding pocket, suggestive of an induced fit mechanism for ligand binding. Curiously, a completely buried ethylene glycol molecule mimics the lactam ring and, thus, serves as a surrogate ligand. The detailed structural delineation of this quorum-quenching antibody will now aid in further development of an antibody-based therapy against bacterial pathogens by interference with quorum sensing.

Introduction

Quorum sensing is a sophisticated system for coordinated regulation of gene expression via cell-to-cell communication in single-cell microorganisms. The term “quorum sensing” refers to the dependence of the signaling activity on the population density and provides bacteria with a means to act as a multicellular unit. Cooperativity in gene expression increases the effectiveness of processes, such as biofilm formation, sporulation, competence, conjugation, virulence factor expression, antibiotic production, swarming motility, or bioluminescence ^{1; 2}. The bacterial population density is sensed by detection of the local concentration of soluble, small hormone-like signaling molecules, also known as autoinducers or quormones, which are constitutively produced at a low basal level. Many Gram-negative bacteria produce N-acyl homoserine lactones (AHLs), whereas Gram-positive species primarily employ peptides or peptide derivatives ². The interspecies signaling molecule autoinducer-2 (AI-2), a furanose derivative, is utilized by both Gram-positive and Gram–negative bacteria. Two proteins are key components of an AHL-based quorum sensing circuit: an autoinducer receptor (R protein), which functions as a transcriptional activator, and an autoinducer synthase (I protein), whose gene expression, among others, is often activated by the autoinducer receptor in a positive feedback loop. As the autoinducer concentration rises as a function of increasing cell-population density, the detection of a minimal threshold stimulatory concentration of the autoinducer leads to alteration of gene expression and, consequently, to a physiological response.

Quorum sensing is crucial for virulence in many pathogenic bacteria that pose significant medical and agricultural threats. In the opportunistic bacterium Pseudomonas aeruginosa, quorum sensing signaling controls the expression of several hundred genes, constituting about 6% of the genome ³. One of these regulated processes is the production of biofilms that are associated with a variety of chronic infections ⁴. Biofilms consist of sessile bacterial colonies encased in polysaccharide matrices that have been shown to be resistant to antimicrobials and host immune cells; hence, treatment of infected cystic fibrosis patients and immune-compromised individuals has proven difficult. Moreover, AHLs themselves induce biochemical changes and exert cytotoxicity in mammalian cells, which further highlights their importance in pathogenicity of Gram-negative bacteria and the need for effective therapeutic countermeasures ^{5; 6}. Thus, strategies to interfere with quorum sensing provide new avenues to combating bacterial diseases in humans, animals, and plants. In fact, several antagonistic approaches, such as heterologous overexpression of quorum-quenching lactonases or discovery of inhibitors against the I or R proteins using combinatorial chemistry, have recently been reported ^{7; 8; 9; 10}. Interference with quorum sensing affords the great benefit of controlling infectious bacteria without interfering with growth, thus avoiding the type of selection pressure that frequently results in development and selection of resistant bacterial strains to antibiotics when traditional antibiotic treatments are used ¹¹.

A completely different approach to quorum quenching has recently been described by harnessing the immune system to counteract the quorum sensing system of the opportunistic pathogen P. aeruginosa ^{12; 13}. In one case, immunization of mice with an AHL-protein conjugate appears to prevent lethality in a P. aeruginosa infection model ¹³. However, AHLs are inherently instable at physiological pH and lead to breakdown products through formation of a ring-opened hydrolyzed form of the AHL, as well as the generation of a tetramic acid species ¹⁴. Thus, in another immunopharmacotherapeutic strategy, the lactone ring of the hapten was replaced with the more stable lactam moiety ¹². Subsequent immunization of mice with three different lactam haptens, RS1, RS2, and RS3 (Figure 1), that closely resemble the two AHLs of P. aeruginosa, N-butanoyl-homoserine lactone (C₄-AHL, 1), and N-(3-oxododecanoyl)-homoserine lactone (3-oxo-C₁₂-AHL, 2), yielded numerous monoclonal antibodies (mAbs) ¹². Affinity measurements using competition ELISA revealed six RS2-antibodies with submicromolar affinity, ranging from 150 to 800 nM, for the 3-oxo-C₁₂-lactam 3, while none of the RS1- or RS3-antibodies possesses an affinity below 10 μM. Consistent with the well-documented specificity of antibodies for their antigens, five of the six RS2-antibodies recognize the original lactam immunogen significantly better than the corresponding lactone 2. For example, antibody RS2-1A4 approximately affords an impressive 1000-fold discrimination between lactone 2 and lactam 3, although these compounds only differ in a single functional group.

Figure 1.

Structures of the two autoinducers (N-acyl homoserine lactones, AHLs) mediating quorum sensing in P. aeruginosa (1-2), of a lactam analog (3), and of haptens (RS1-3). The affinity constants of RS2-1G9 for 2 and 3 are listed.

However, cross-reactivity with the lactone is explicitly desired in this case, since the P. aeruginosa quorum sensing signaling molecules are indeed lactones, not lactams. Notably, one of the characterized antibodies, termed RS2-1G9, met this requirement and bound lactone 2 with an even higher affinity (K_d = 150 nM) than lactam 3 (300 nM). The use of an analog of the intended target in the immunization process to elicit antibodies against small molecules has been successfully used in the past for the generation of catalytic antibodies, in a so-called “bait-and-switch” strategy ^{15; 16}. Consistent with high affinity recognition of AHL 2, subsequent reporter assays demonstrated that mAb RS2-1G9 effectively inhibits quorum sensing signaling in P. aeruginosa ¹². At the same time, mAb RS2-1G9 was able to discriminate against the closely-related quorum sensing molecule C₄-AHL 1, since this compound is bound with 1,000-fold lower affinity.

In order to gain insight into immune recognition of a quorum sensing molecule by an antibody and advance the development of antibody-based antimicrobial therapeutics that target quorum sensing, crystallographic studies of antibody RS2-1G9 were initiated. From a structural point of view, the generation of an antibody with nanomolar affinity against a lipid-like compound, such as 3, featuring a small head group that lacks typical haptenic features, such as aromaticity or charge, is quite remarkable ^{12; 17}. Moreover, structure determination of a RS2-1G9-ligand complex has also provided a structural basis for its cross-reactivity with lactones and lactams and for its high specificity for the 3-oxododecanoyl substituent in AHLs.

Results and Disscussion

Quality of the RS2-1G9 lactam complex crystal structure

The crystal structure of the Fab fragment of antibody RS2-1G9 in complex with the AHL lactam analog 3 was determined by molecular replacement and refined to 3.18 Å resolution. The bound ligand has three additional methylene units in the acyl chain and lacks the terminal carboxyl group in comparison to the immunizing hapten RS2 (Figure 1). Despite the modest resolution, R_cryst and R_free are better than average for this resolution range (R_cryst = 21.0% and R_free = 26.5%, Table 1) ¹⁸. Only Thr^L51 in the complementarity-determining region (CDR) L2 of both Fab molecules in the asymmetric unit are in the “disallowed” region of the Ramachandran plot, but both have well-defined electron density. Thr^L51 is in a γ turn, as commonly observed in the canonical CDR L2 structure of other antibody structures¹⁹ and, hence, is not a true outlier despite the Procheck designation.

Table 1.

Data collection and refinement statistics of RS2-1G9 crystal structures

	AHL analog complex	ethylene glycol complex
Space group	P43212	P212121
Unit cell dimensions (Å)	a=b=118.7, c=176.0	a=56.3, b=72.0, c=116.1
Resolution range (Å)	50.0-3.18 (3.25-3.18)	50.0-2.85 (2.92-2.85)
Unique reflections	19,140	11,938
Completeness (%)	88.6 (95.0)	99.6 (99.9)
Redundancy	2.2 (2.1)	3.6 (3.6)
Wilson B-value (Å2)	79.9	83.5
Rsym	0.096 (0.508)	0.062 (0.534)
<l/σ>	14.2 (2.0)	26.1 (2.7)
Rcryst/ Rfree	0.210/ 0.265	0.200/ 0.263
Fabs in asymmetric unit	2	1
Rmsd from ideal bond lengths (Å)/ angles (°)	0.014/ 1.5	0.013/ 1.5
Average B-values protein/ligand (Å2)	75.3/ 59.0	60.3/ 58.8
Ramachandran plot most favored/ additionally allowed/ generously allowed/ disallowed (%)	83.2/15.1/1.4/0.3	84.9/14.6/0.3/0.3

* Highest resolution shell.

^† ${\mathrm{R}}_{\mathrm{sym}}={\sum }_{\mathrm{hkl}}{\sum }_{\mathrm{i}}\mid {\mathrm{I}}_{\mathrm{i}}\left(\mathrm{hkl}\right)-<{\mathrm{I}}_{\mathrm{i}}\left(\mathrm{hkl}\right)>\mid /{\sum }_{\mathrm{hkl}}{\sum }_{\mathrm{i}}\mid {\mathrm{I}}_{\mathrm{i}}\left(\mathrm{hkl}\right)\mid$

^‡ ${\mathrm{R}}_{\mathrm{cryst}}={\sum }_{\mathrm{hkl}}\mid \mid {\mathrm{F}}_{\mathrm{c}}\left(\mathrm{hkl}\right)\mid -\mid {\mathrm{F}}_{\mathrm{o}}\left(\mathrm{hkl}\right)\mid \mid /{\sum }_{\mathrm{hkl}}\mid {\mathrm{F}}_{\mathrm{o}}\left(\mathrm{hkl}\right)\mid$

^§ R_free is calculated as for R_cryst, but from 5% of the data that was not used for refinement.

^¶ Root-mean-square deviation.

^∥ Thr^L51 (CDR L2) is the only residues in a disallowed region, but Thr^L51 has well-defined electron density and is in a γ turn, as commonly observed in other antibody structures.

Overall, the electron density was of good quality and did not show any main-chain breaks at a contour level of 1σ throughout the entire structure. Even the notoriously poorly defined loop between Ser^H127 and Gln^H135 of the heavy-chain constant domain C_H1 was visible in the electron density map ^{20; 21}, albeit exhibiting reduced electron density. Most importantly, clear electron density at the antibody combining sites in both Fabs became evident in σ_a-weighted 2F_o–F_c and F_o–F_c maps during refinement (Figure 2), which guided the incorporation of the lactam ligand into the structure. The terminal part of the acyl chain projects into solution and, accordingly, is less well-defined in the electron density map. The ligand refined well, as evidenced by a slightly lower average B-value calculated over all ligand atoms with respect to that of the protein (Table 1).

Figure 2.

Antibody combining site of RS2-1G9 bound to an AHL lactam mimetic (pink). The light and heavy chains are colored in yellow and blue, respectively. The σ_a-weighted 2F_o-F_c electron density map around the ligand is contoured at 1.4σ. The microenvironment at the very bottom of the binding pocket tolerates high-affinity binding of both the lactam and the lactone, since Asn^L34 does not form a hydrogen bond to the NH-group of the lactam. A potential hydrogen bond in the lactone complex model with the NH₂-group of Asn^L34 may account for its increased affinity with respect to the lactam. The displayed orientation of the terminal amide group of Asn^L34 is preferred due to formation of a hydrogen bond with Tyr^L32. CDR L3 is omitted for clarity.

Overall Fab structure and architecture of the RS2-1G9 combining site

The structure of Fab RS2-1G9 strongly resembles other Fab molecules in its overall topology and features ²². Notably, the two copies of RS2-1G9 exhibit large elbow angles of 211° and 210°, respectively, consistent with RS2-1G9 possessing a λ light chain ²³. The hapten analog 3 is bound in the center of the antibody combining site (Figure 3). As observed in other antibody complexes with small ligands ²⁴, CDR H3 and L3 are primarily responsible for ligand recognition (43% and 19% of the total Fab surface area ²⁵ contacting 3, respectively), H1 (16%), L1 (13%), and H2 (9%) make fewer contacts, whereas L2 does not contribute to ligand binding. Most striking, however, is the complete encapsulation of the lactam head group within the antibody combining site of RS2-1G9 (Figure 4), burying 97% of the surface area of the lactam ring and the first six carbon units of the C₁₂ acyl chain. A narrow tunnel linking the lactam-binding cavity with bulk solvent harbors half of the acyl tail. Unlike many other anti-hapten antibodies that also provide deep, but rigid, hydrophobic pockets or slots ^{24; 26}, mAb RS2-1G9 features a narrow constriction within the tunnel that envelopes the linkage between the acyl chain and the lactam. As a consequence of this architecture, facile diffusion of the ligand into and out of the binding site is not possible without major conformational changes. Among the few antibody-ligand structures that similarly clasp their ligands in deep, constricted cavities, are the catalytic antibodies 4C6 (95%), 1F7 (90%), or the Diels-Alder antibody 13G5 (99%) ^{27; 28; 29}.

Figure 3.

Architecture of the antibody combining site of RS2-1G9. The CDRs of light and heavy chains are highlighted in yellow and blue, respectively. The lactam 3 is shown in pink. The tip of the CDR H3 loop bends over the ligand and largely seals it from bulk solvent.

Figure 4.

High electrostatic and shape complementarity of the hapten analog in the antibody-combining site. A slice through the center of the binding site is shown. (a) and (b) correspond to the front and back view. The electrostatic potential was calculated in APBS ⁶⁴ and mapped onto the surface with the color code ranging from -30 kT/e (bright red) to +30 kT/e (dark blue).

In addition to complete burial of the ligand head group, RS2-1G9 also exhibits high shape complementarity (S_c parameter of 0.84 for the head group including the 3-oxo-group, 0.81 for the whole ligand) to the AHL lactam analog 3 ³⁰, thus ranking among antibody-hapten complexes which exhibit high S_c values, such as antibody 34E4 (S_c = 0.87) or 7A1 (S_c = 0.89) ^{31; 32}. It is conceivable that the shape correlation in RS2-1G9 may even be higher, since no water molecules could be detected in the electron density due to the modest resolution. Three water molecules considerably contributed to the high S_c parameter in the antibody-hapten complex of 34E4 ³¹.

AHL lactam analog recognition in antibody RS2-1G9

Ligand recognition in mAb RS2-1G9 is realized by a combination of 85 van der Waals’ contacts and six hydrogen bond interactions with the functional groups of the lactam analog (Figure 5) ³³. Specifically, the aromatic side chains of Tyr^L32, Trp^L91, Trp^L96, Trp^H33, Tyr^H58, and Phe^H100E snugly nestle around the ligand and provide 47% of the buried surface area of the Fab binding pocket. With 33Ų each, Trp^L91 and Trp^L33 contribute most to this surface area. Interestingly, Trp^L91 is the hallmark residue of CDR L3 in λ chains and its central role in ligand recognition observed here further underlines the common gene usage of λ chains in antibodies generated against various small molecules. For instance, the indole ring of Trp^L91 is buttressed against the ligands in the complex structures of the anti-benzimidazolium antibody 34E4 ³¹, the anti-nitrophenyl antibodies SPE7 ³⁴, N1G9 ³⁵, 88C6/12 ²⁰, the anti-carbohydrate antibody Se155-4 ³⁶, and the anti-indium(III)-benzyl-EDTA antibody CHA255 ³⁷, to cite just a few examples. Furthermore, in most of these antibodies, as well as in RS2-1G9 (Figure 5), the conserved Tyr^L32 and Trp^L96 in the immediate vicinity of Trp^L91 further complement its ligand interactions.

Figure 5.

Antibody combining site of RS2-1G9 (stereoview). Hydrogen bonds are shown as broken lines. Only Fab side chains that contact the lactam ligand (pink) are displayed. The hapten analog satisfies all its functional groups, except for the amide group of the lactam, which is crucial for the observed cross-reactivity of this antibody with an N-acyl homoserine lactone. A plethora of aromatic side chains surrounds the ligand. The residues between Leu^H97 and Asn^H100 of CDR H3 are omitted for clarity.

The engulfment of the ligand is structurally accomplished by a 90° bend of the tip of the CDR H3 loop over the ligand (Figure 3). Surprisingly, CDR H3 contributes more ligand contacts via main-chain (77) than via side-chain atoms (69), which is quite unusual for hapten recognition. The only other main-chain contacts are provided by Ser^L93 (6) in CDR L3. Of the CDR H3 residues, Phe^H100E (16) at the base of the binding pocket and Asn^H100A (31) at the tip of CDR H3 play major roles in ligand binding. In addition, the side-chain carbonyl of the Asn^H100A specifically interacts with the ligand by engaging in a hydrogen bond with the exocyclic amide (3.0 Å). The adjacent carbonyl group of the acyl tail undergoes simultaneous hydrogen bonding interactions with the backbone amide of Ser^H96 (2.9 Å) and Nε2 of His^H35 (2.6 Å). The 3-oxo group of the acyl chain hydrogen bonds with NHε1 of Trp^L96 (3.1 Å). Lastly, the carbonyl oxygen of the lactam ring makes hydrogen bonds to the backbone amide of Gly^100C (2.8 Å) and the hydroxyl of Ser^H96 (2.9 Å). Most of these hydrogen bonds display optimal geometry, as assessed by the program HBPLUS ³⁸, thereby mediating strong specific interactions and providing convincing evidence for correct orientation of the ligand within the antibody-hapten complex despite the modest resolution of the data.

Structural basis for discrimination of RS2-1G9 against C₄-AHL

The mAb RS2-1G9 shows high affinity for 3-oxo-C₁₂-AHL (2), while the other quorum sensing molecule of P. aeruginosa, C₄-AHL (1), is recognized with 1000-fold lower affinity ¹². Based on the crystal structure, the missing 3-oxo group of AHL 1 would lead to the loss of a strong hydrogen bond interaction with NHε1 of Trp^L96. Furthermore, the lack of eight methylene units in the acyl chain of C₄-AHL would abrogate 11 van der Waals’ contacts that are observed in the 3-oxo-C₁₂-lactam complex structure. Therefore, the combination of increased van der Waals’ contacts and the additional hydrogen bond appears to be responsible for the high specificity of RS2-1G9 for 3-oxo-C₁₂-AHL vs. C₄-AHL.

3-oxo-C₁₂-AHL recognition: structural basis for cross-reactivity

Analysis of the RS2-1G9 ligand interactions reveals that the amide group of the lactam ring is the only functional group without satisfied hydrogen bonding potential and suggests that the chemical nature of this position is less critical for ligand recognition. The corresponding lactone featuring an oxygen atom at this location can easily be accommodated in place of the lactam without any detrimental effect on high-affinity binding, thus providing a structural basis for its cross-reactivity with lactone 2. Strikingly, the 3-oxo-C₁₂-AHL 2 exhibits about twofold higher affinity for RS2-1G9 than the corresponding lactam 3 ¹². Detailed analysis of this part of the binding pocket can offer a plausible explanation for this binding behavior. The side-chain oxygen and nitrogen of Asn^L34 are both in hydrogen-bonding distance to the lactam NH-group (2.9 Å and 3.3 Å, respectively). However, the lactam is unable to form a hydrogen bond with the Asn^L34 carbonyl due to unfavorable geometry (Figures 2 and 5), as assessed by the program HBPLUS ³⁸. By contrast, if the lactam NH group is substituted with O, the corresponding analysis of this lactone-complex model does, indeed, reveal possible formation of a hydrogen bond between the lactone oxygen and the NH₂–group of Asn^L34, albeit with suboptimal geometry. As this interpretation critically depends on the proper orientation of the Asn^L34 carboxamide, the final RS2-1G9 structure was submitted to an all-atom contact analysis performed by the program MolProbity ³⁹ with the terminal carboxamide group of Asn^L34 in one of the two possible orientations and flipped by 180° in the other calculation. The analysis provided clear evidence for the rotamer displayed in Figure 2, which is favored by formation of a hydrogen bond of its NH₂-group with the backbone carbonyl of Tyr^H32 (2.9 Å)(Figures 2 and 5). In summary, the microenvironment at the very bottom of the binding pocket tolerates high-affinity binding of both the lactam and the lactone without much discrimination; slightly improved hydrogen-bonding potential in the antibody-lactone complex may account for its increased affinity with respect to the lactam.

CDR H3 has uniquely evolved for high-affinity recognition of its immunogen

The successful selection of only a small number of RS2-antibodies with affinities in the submicromolar range from 68 mAbs illustrates the difficulties in generating high-affinity antibodies against small lipophilic structures ¹². The crystal structure of one representative of these, RS2-1G9, in complex with its hapten has revealed that the attributes of high shape complementarity, deep burial, hydrogen bonding to all but one functional group of the ligand, and numerous van der Waals’ contacts can explain the high affinity of RS2-1G9 towards specific N-acyl homoserine lactones and lactam analogs. In particular, hapten recognition is primarily achieved by simultaneously optimizing shape complementarity and hydrogen bonding of CDR H3. In fact, this hypervariable loop has been found to dominate immune recognition in most antibody-hapten complexes and to be the most flexible and diverse CDR loop ^{24; 40; 41}. Although the conformation of CDR H3 is somewhat correlated with length, the backbone conformation of CDR H3 is unique to RS2-1G9 and does not superimpose with those of other antibodies containing the same CDR H3 length , such as the anti-digoxin antibody 40-50, the anti-phospholipase C_δ1 antibody L5MK16 ⁴², or the anti-neuraminidase antibody NC10 ⁴³. With five insertions after residue Asn^H100, antibody RS2-1G9 has a relatively long CDR H3 loop for mouse antibodies. On average, the length of CDR H3 in murine anti-hapten antibodies is only 8.5 residues, while RS2-1G9 consists of 13 amino acids ⁴⁰. Again, the crystal structure has illuminated the structural basis for the origin of the relatively long CDR H3 loop, as it forms a flap at its tip in order to make maximal interactions with the lactam ligand.

Crystal structure of RS2-1G9 in complex with ethylene glycol

The constriction of the antibody binding pocket harboring part of the acyl chain is too narrow to allow facile entrance or exit of the ligand head group. As a consequence, structural rearrangements of the antibody combining site must occur upon ligand binding and release. In order to experimentally address this issue, we crystallized RS2-1G9 in absence of its cognate lactam or lactone ligands. Notably, these crystals were indexed in a different space group than those of the lactam complex (Table 1), possibly indicating structural changes. However, the antibody in the crystal was not in its truly unliganded form, as ethylene glycol, which most likely originated from the cyroprotection buffer containing 25% ethylene glycol, occupied the region of the binding pocket that harbors the lactam moiety in the complex structure (Figure 6). This Fab structure was determined to 2.85 Å resolution and refined to an R_cryst of 20.0% and an R_free of 26.3% (Table 1). Although the resolution of the ethylene glycol-bound RS2-1G9 structure is better than that of the lactam complex (3.18 Å), the framework regions of the variable part, as defined by Kabat and Wu ⁴⁴, as well as the constant domains are essentially identical to its lactam complex with an rmsd of 0.6 Å for the Cα atoms, which validates the conclusions from the lower resolution structure of the RS2-1G9 lactam complex.

Figure 6.

Ethylene glycol acts as a surrogate ligand in the “unliganded” antibody RS2-1G9 (lactam complex in grey, ethylene glycol complex in green, stereoview). This solvent molecule (yellow) substitutes for the lactam ring of the bound hapten derivative (pink) and fills the bottom of the ligand cavity. Strikingly, the bound ethylene glycol adopts an unfavorable, nearly eclipsed conformation that faithfully mimics part of the lactam ring. Ethylene glycol was added for cryoprotection of the crystals prior to data collection at cryogenic temperatures. This view also illustrates how CDRs H3 and L3 rearrange to close the N-acyl harboring tunnel of the binding site.

Antibody RS2-1G9 binds its cognate ligands via an induced fit mechanism

In contrast to the framework regions, the architecture of the RS2-1G9 combining site has significantly changed in absence of its bound lactam derivative. Importantly, the main chain, as well as side-chains, of the CDR loops have clear and distinctly different electron densities in both complexes. While the complementarity determining region of the lactam complex structure features a deep binding cavity, this region has transformed into a relatively flat surface in the ethylene glycol-bound form (Figure 7). The structural changes in CDRs H3 and L3 are primarily responsible for this remodeling. Instead of rigid-body movements of the H3 and L3 tips, individual residues undergo disparate structural changes that collectively lead to closure of the tunnel harboring the N-acyl chain in the hapten derivative complex (Figures 6 and 8). In quantitative terms, H3 undergoes the largest rearrangement (rmsd of 1.6Å for its Cα atoms), followed by L3 (0.9Å) (Figure 8). By contrast L1, L2, H2, and H1 do not undergo substantial movements: 0.7Å, 0.6Å, 0.5Å, and 0.3Å, respectively. The structural repositioning of the H3 loop is accompanied by even more pronounced reorientations of its side-chains (Figure 8). Asn^H100A is displaced from the center to the edge of the antibody combining site in the ethylene glycol-bound structure. Conversely, Asn^H100 and Tyr^H99 move towards the center and seal off the binding pocket via their side chains. By packing against the indole ring of Trp^L91 and the phenol ring of Tyr^L32, Asn^H100 partly covers their hydrophobic surfaces. Likewise, Tyr^H99 largely restricts solvent accessibility to Trp^H33 and Tyr^H58. Finally, Asn^L94 of L3 wedges between Tyr^H58 and Trp^L91 and completes the formation of the flat molecular surface of the RS2-1G9 combining site in the “unliganded” state. On average, the side-chains atoms of Tyr^H98, Tyr^H99, Asn^H100, Asn^H100A, and Asn^L94 (labeled green in Figure 8) have an rmsd of 5.6 Å in the two structures, highlighting the large structural plasticity in the antigen binding site of RS2-1G9.

Figure 7.

Comparison of the molecular surface representation of (a) the lactam complex and (b) the ethylene glycol complex of antibody RS2-1G9 reveals profound differences in the architecture of the antibody combining site. The lactam and the buried ethylene glycol (EG) are colored in pink and yellow, respectively.

Figure 8.

Induced fit in RS2-1G9 (lactam complex in grey, ethylene glycol complex in green, stereoview). Upon ligand binding, the main-chain and side-chain atoms of CDR H3 undergo the largest rearrangements, while L3, L1, and H2 move to a minor extent. L2 and H1 are essentially not in contact to the hapten derivative and, hence, are not displayed. In particular, the tip of H3 completely reorganizes upon ligand binding to accommodate the lactam in the binding pocket (induced fit). For clarity, Tyr^H98, Tyr^H99, Asn^H100, and Asn^H100A are labeled in both structures.

Ethylene glycol serves as a surrogate ligand

Although the antibody combining site reveals considerable structural rearrangements at the tips of the CDR loops, the buried chamber that is occupied by the lactam moiety in the RS2-1G9-3 complex (Figure 4) is not subject to structural changes, but harbors ethylene glycol (Figure 6). Strikingly, ethylene glycol adopts an energetically unfavorable eclipsed conformation, thus somewhat mimicking the five-membered lactam ring (Figure 6). Precedence for antigen binding pockets that are occupied by molecules of the crystallization buffer in absence of their cognate haptens has been documented previously ^{27; 32}. Moreover, closure of the ligand binding pocket in RS2-1G9 by antibody-derived residues that substitute for high-affinity ligands has also been observed in other anti-hapten antibodies ^{45; 46; 47; 48; 49}. In the anti-progesterone antibody DB3, for instance, a tryptophan side chain partially occupies the steroid binding cavity in the free antibody and acts as an antibody-derived surrogate ligand to “solvate” an otherwise hydrophobic site in the absence of antigen ^{45; 46}. Thus, partial closure of the highly-hydrophobic pockets in DB3 and RS2-1G9 appears to be thermodynamically driven, since the structural alteration reduces the percentage of non-polar residues exposed to the aqueous environment in the unliganded state.

Structure-based protein engineering of the quorum-quenching antibody RS2-1G9

The crystal structures of the quorum-quenching antibody RS2-1G9 provide an excellent starting point for protein engineering in order to increase the therapeutic potential of this antibody in its interference with quorum sensing of the human pathogen P. aeruginosa. In particular, site-directed mutagenesis is a powerful tool to fine-tune and to improve interactions of the antibody with AHL. Based on their vicinity to the AHL lactam analog, the following mutations are likely to influence ligand-antibody interactions in RS2-1G9: His^H35→Gln, Tyr^H58→Phe/Trp, Asn^H100A→His/Gln, Phe^H100E→His/Trp, and Asn^L34→Asp/Glu/His/Lys/Gln. Since these mutations are predominantly conservative, improved packing or hydrogen bonding to 3-oxo-C₁₂-AHL may result from these substitutions. Moreover, the mutation of Asn^L34 might introduce lactonase activity into the antibody due its juxtaposition to the ring carbonyl carbon, which is hydrolytically attacked in naturally-occurring quorum-quenching enzymes ^{50; 51}. The introduction of catalytic activity would greatly enhance the potency of antibody RS2-1G9, since a single antibody molecule would degrade many AHLs, thus requiring significantly lower titers of antibody for effective treatment in vivo ^{52; 53}. Importantly, mutagenesis may alternatively result in a covalent binding event between the RS2-1G9 mutants and 3-oxo-C₁₂-AHL rather than catalysis, as it is conceivable that a lactone bond breaking event might take place without the subsequent replacement with a water molecule and release of the hydrolyzed 3-oxo-C₁₂-AHL. Previously, such covalent catalysis has been achieved in catalytic aldolase antibodies that feature enzyme-like reaction rates and broad substrate specificity ^{54; 55}. Finally, biochemical characterization of the mutations will complement our crystallographic characterization of ligand recognition by RS2-1G9 and corroborate key residues in the interaction between the antibody and AHL.

Conclusions

The crystal structure of antibody RS2-1G9 in complex with an AHL analog clearly illustrates how the immune system has evolved to potently bind and discriminate a low molecular weight lipophilic compound with a small head group, such as the AHL lactam analog 3, with nanomolar affinity. The head group is deeply buried within the protein and specifically recognized by hydrogen bonds to the functional groups. The tunnel towards the protein surface as well as the potential dynamics of the CDR H3 loop enable mAb RS2-1G9 to reversibly bind to its ligands. Evidence for substantial structural plasticity in the RS2-1G9 combining site was gleaned from the crystal structure of the free Fab in the absence of any cognate ligands. Although the antibody already achieves great shape and electrostatic complementarity towards the AHL mimic, further optimization of antibody-ligand affinity or introduction of lactonase activity by site-directed mutagenesis may represent promising routes to enhanced therapeutic potential of the quorum-quenching antibody RS2-1G9. Improved hapten design and the synthesis of new haptens specifically designed to emulate AHLs of bacteria other than P. aeruginosa may yield additional more potent antibodies with the desired properties of quorum quenching.

Materials and Methods

Fab preparation, crystallization, and structure determination of Fab RS2-1G9 in complex with an AHL lactam analog and with ethylene glycol

The Fab fragment was produced from murine mAb RS2-1G9 (IgG1, λ) by pepsin digest for 4 hours using standard protocols ⁵⁶. Crystallization experiments were performed by the sitting drop vapor diffusion method at 22.5°C. The Fab, concentrated to 15mg/ml in 0.1M sodium acetate pH 5.5, was crystallized after several weeks in presence of 4-fold molar excess of lactam 3 from 1.0M Na/K tartrate, 0.2M NaCl, and 0.1M imidazole buffer pH 8.0. The ethylene-glycol bound RS2-1G9 (“unliganded”) was crystallized from 10% PEG 8000, and 0.1 M imidazole pH 6.5. For data collection, the crystals were flash cooled to 100K using 25% glycerol (lactam complex) and 25% ethylene glycol (ethylene glycol complex) as a cryoprotectant, respectively. Lactam-complexed Fab and “unliganded” Fab data were collected at synchrotron beamlines APS 23-ID-D and SSRL 11-1, respectively, and processed and scaled with HKL2000 ⁵⁷ (Table 1). The structure of the RS2-1G9-3 complex was determined by molecular replacement using the program Phaser ⁵⁸ and the coordinates of scFv B1-8 (PDB ID code 1A6V) for the variable domain and Se155-4 (1MFB) for the constant domain. The structure of RS2-1G9 in complex with ethylene glycol was determined by molecular replacement using the same program and the RS2-1G9 coordinates of the lactam complex. The models were refined by alternating cycles of model building with the program O ⁵⁹ and refinement with Refmac5 ⁶⁰. Given the modest resolution of 3.18 Å and 2.85 Å, tight non-crystallographic symmetry restraints were applied. The final statistics are shown in Table 1. The quality of the structures was analyzed using the programs MolProbity ³⁹, WHAT IF ⁶¹, and PROCHECK ⁶². All figures were prepared with PyMol ⁶³. The coordinates and the structure factors are deposited at the PDB under accession code 2NTF (RS2-1G9-3 complex) and 2OP4 (RS2-1G9 ethylene glycol complex), respectively.