Structural basis for effectiveness of siderophore-conjugated monocarbams against clinically relevantstrains of Pseudomonas aeruginosaSeungil Han1, Richard P. Zaniewski, Eric S. Marr, Brian M. Lacey, Andrew P. Tomaras, Artem Evdokimov2,J. Richard Miller3, and Veerabahu Shanmugasundaram

Pfizer Global Research and Development, Eastern Point Road, Groton, CT 06340

Edited* by E. J. Corey, Harvard University, Cambridge, MA, and approved October 25, 2010 (received for review September 1, 2010)

Pseudomonas aeruginosa is an opportunistic Gram-negative patho-gen that causes nosocomial infections for which there are limitedtreatment options. Penicillin-binding protein PBP3, a key therapeu-tic target, is an essential enzyme responsible for the final steps ofpeptidoglycan synthesis and is covalently inactivated by β-lactamantibiotics. Herewedisclose the first high resolution cocrystal struc-tures of the P. aeruginosa PBP3with both novel andmarketed β-lac-tams. These structures reveal a conformational rearrangement ofTyr532 and Phe533 and a ligand-induced conformational change ofTyr409 and Arg489. The well-known affinity of the monobactamaztreonam for P. aeruginosa PBP3 is due to a distinct hydrophobicaromatic wall composed of Tyr503, Tyr532, and Phe533 interactingwith the gem-dimethyl group. The structure of MC-1, a new sidero-phore-conjugated monocarbam complexed with PBP3 providesmolecular insights for leadoptimization. Importantly,wehave iden-tified a novel conformation that is distinct to the high-molecular-weight class B PBP subfamily, which is identifiable by common fea-tures such as a hydrophobic aromaticwall formedby Tyr503, Tyr532,and Phe533 and the structural flexibility of Tyr409 flanked by twoglycine residues. This is also the first example of a siderophore-conjugated triazolone-linked monocarbam complexed with anyPBP. Energetic analysis of tightly and loosely held computed hydra-tion sites indicates protein desolvation effects contribute signifi-cantly to PBP3 binding, and analysis of hydration site energiesallows rank ordering of the second-order acylation rate constants.Taken together, these structural, biochemical, and computationalstudies provide a molecular basis for recognition of P. aeruginosaPBP3 and open avenues for future design of inhibitors of this classof PBPs.

antibiotic resistance ∣ cell wall ∣ transpeptidase ∣ covalent inhibitor

Infections caused by Gram-negative pathogens are a seriousthreat to public health. Multidrug-resistant Gram-negative or-

ganisms such as Pseudomonas aeruginosa, Klebsiella pneumoniae,Escherichia coli, and Acinetobacter baumannii are emerging assignificant pathogens. Therapeutic options for their treatmentare limited, in part due to more focused efforts over recent yearsto combat Gram-positive bacteria, such as Staphylococcus aureusand vancomycin-resistant enterococci (1, 2). Simultaneously, theemergence and spread of resistance mechanisms including β-lac-tamases have diminished the value of many marketed β-lactamantibiotics. New Gram-negative antibiotics for the treatment ofnosocomial infections are urgently needed to tackle the highmorbidity and mortality rates associated with multidrug resistantpathogens.

P. aeruginosa causes multiple types of infections includingpneumonia, bacteremia, and urinary tract, ear, skin, and softtissue infections. Current therapies for P. aeruginosa infectionsinclude broad-spectrum β-lactam antibiotics, carbapenems suchas imipenem or meropenem, or a more Gram-negative selectivemonobactam such as aztreonam (3). In addition to producingβ-lactamases, P. aeruginosa can exhibit or acquire additional re-

sistance mechanisms including reduction of outer membrane per-meability, expression of efflux pumps, or by mutating the targetpenicillin-binding proteins (4).

The macromolecular targets for β-lactam antibiotics are peni-cillin-binding proteins (PBPs), membrane-associated periplasmicenzymes that catalyze essential transpeptidase reactions involvingpeptidoglycan, a major component of bacterial cell walls. At leastsix PBPs of P. aeruginosa have been detected by their ability toform covalent adducts with radiolabeled penicillin G (5). Basedon their structural features and enzyme activities, P. aeruginosaPBPs are categorized into three classes. High-molecular-weightclass A PBPs (PBP1a and PBP1b) are bifunctional enzymes con-taining both transglycosylase and transpeptidase activities. High-molecular-weight class B PBP, of which Pseudomonas PBP3 is amember, possesses only transpeptidase activity. Low-molecular-weight class C PBPs, including PBP4 and PBP5, are soluble pro-teins and act as DD carboxypeptidases. The high-molecular PBPsare essential for cell growth and are validated and precedentedtargets for β-lactam antibiotics (6). PBP3, encoded by the ftsIgene, is the only known peptidoglycan synthase in E. coli and isessential for cell division. PBP3 inhibition by a β-lactam and genedeletion inhibit cell septation, resulting in filamentation (7). PBP2from Neisseria gonorrhoeae (NgPBP2), is the only other knownPBP3 ortholog from a Gram-negative pathogen that has beencrystallized. It has two flexible loop regions that appear to blockentry of inhibitors into the active site, resulting in unsuccessfulsoaking experiments despite numerous attempts to produceNgPBP2-inhibitor complexes (8).

One approach to circumvent permeability-mediated resistanceof Gram-negative pathogens such as P. aeruginosa is to utilize thesiderophore-mediated iron acquisition system to effectively deli-ver drug molecules to the periplasmic space (9). Bacteria requireiron in order to survive, and the acquisition of iron depends onthe production, release, and active reuptake of iron-scavengingmolecules called siderophores. Our strategy relies on the incor-poration of a hydroxypyridone moiety as a siderophore attached

Author contributions: S.H., R.P.Z., B.M.L., A.P.T., and V.S. designed research; S.H., R.P.Z.,E.S.M., B.M.L., A.E., J.R.M., and V.S. performed research; R.P.Z., E.S.M., B.M.L., A.E., andJ.R.M. contributed new reagents/analytic tools; S.H., A.P.T., J.R.M., and V.S. analyzed data;and S.H. and V.S. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Data deposition: The crystallography, atomic coordinates, and structure factors have beendeposited in the Protein Data Bank, www.pdb.org (PDB ID codes 3PBN, 3PBO, 3PBQ, 3PBR,3PBS, and 3PBT).1To whom correspondence should be addressed. E-mail: seungil.han@pfizer.com.2Present address: Structural Biology, Monsanto Company, 700 Chesterfield Parkway West,Chesterfield, MO 63017.

3Present address: Department of In Vitro Pharmacology, Merck Research Labs, 33 AvenueLouis Pasteur, Boston, MA 02115.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1013092107/-/DCSupplemental.

22002–22007 ∣ PNAS ∣ December 21, 2010 ∣ vol. 107 ∣ no. 51 www.pnas.org/cgi/doi/10.1073/pnas.1013092107

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1

to a monocyclic β-lactam antibiotic, which is anticipated to resultin increased concentrations of drug in the periplasm (10).

To help elucidate the molecular mechanism of P. aeruginosaPBP3 (PaPBP3) and its β-lactam recognition elements, we solvedmultiple X-ray structures of a soluble form of PaPBP3. High-resolution structures of the apo and ceftazidime-bound formswere obtained as well as complexes with meropenem, imipenem,and aztreonam. Extensive medicinal chemistry efforts led to theidentification of MC-1, a siderophore-conjugated monocarbamthat utilizes a carbonylaminosulfonyl species for β-lactam activa-tion, and which shows remarkable antipseudomonal activity bothin vitro and in vivo. The crystal structure of PaPBP3 complexedwith MC-1 at 1.64-Å resolution reveals a unique binding mode ofthe siderophore. WaterMap analysis was carried out to evaluatethe correlation between protein desolvation energies and thesecond-order acylation constants (k2∕Kd) during ligand binding.A ligand-dependent water displacement effectiveness score (ϵi)provides a unique measure that rank orders the second-orderacylation rate constants. Our biochemical, computational, andstructural results provide key insight into recognition elementsand avenues to advance the design of β-lactam inhibitors with im-proved efficiency of PBP3 interaction and effectiveness againstGram-negative pathogens.

Results and DiscussionOverall Structures of PaPBP3. PaPBP3 has overall dimensions of∼45 × 63 × 100 Å and consists of two domains, an N-terminalnonpenicillin-binding domain and a C-terminal transpeptidase(TP) domain (Fig. 1A). The N-terminal domain, resembling apair of sugar tongs, extends with long β-strands ∼45 Å in length,which are unique to the high-molecular mass of PBPs of subclassB3 (11). The four conserved motifs of the N-terminal domain arelocated at the base of the sugar tongs and form important struc-tural elements involved in interdomain interactions (Fig. S1).The N-terminal domain of the high-molecular-weight PBPs isrequired for the folding and stability of the C-terminal transpep-tidase domain and has been suggested to provide recognition sitesfor formation of the multiprotein complex responsible for cellwall biosynthesis (12). The structure of the C-terminal domain(residues 225–579) is similar to other TP domains of PBPs, carry-ing the classical signature of the penicilloyl serine transferasesuperfamily. The active site including the nucleophilic Ser294residue is located between two subdomains, the α-subdomainand α/β-subdomain. The α-subdomain containing α2, α4–α6, andα8 forms one side of the active site groove. A β-hairpin extensionconnecting α2 and α4 is composed of four short β-strands (β2a,β2b, β2c, and β2d) and forms two perpendicular antiparallelβ-sheets (β2a/β2d and β2b/β2c) at the top of the active site.The α/β subdomain contains a central core of a five-stranded anti-parallel β-sheet (β3/β4/β5/β1/β2) sandwiched by several helices onboth sides (α1, α9, α10, and α11). In all six PaPBP3 structures, thetwo loop regions (residues 332–338 connecting β2c to β2d andresidues 526–533 connecting α5 to α11) display substantial con-formational changes upon binding different inhibitor classes andplay an important role in substrate/inhibitor recognition.

Induced-Fit-Conformational Changes of the PaPBP3 in Complex withCeftazidime. Ceftazidime, a third-generation cephalosporin, hasgood in vitro activity against P. aeruginosa and is clinically effec-tive and safe in treating many nosocomial bacterial infections.Ceftazidime is composed of a β-lactam ring fused with a six-mem-bered dihydrothiazine ring, a methyl pyridine at the C3 position, acarboxylic acid group at the C4 position, and a bulky aminothia-zole-containing side chain at the C7 position (Fig. 1B). The crys-tal structure of PaPBP3 in complex with ceftazidime at 1.74-Åresolution clearly identifies ceftazidime covalently bound toSer294 and the rearrangement of the dihydrothiazine ring to forman exocyclic methylene group at C3 position due to the departure

of the pyridine moiety (Fig. S2) (13). The carbonyl oxygen of thecovalent enzyme-ceftazidime ester is positioned into the oxyanionhole and hydrogen bonds with the main chain amides of Ser294and Thr487. The C4 carboxylic acid group is anchored by twohydrogen bonds with the hydroxyl groups of Ser485 and Thr487in the KSGT motif. The Lys297 in the conserved SXXK motif nolonger hydrogen-bonds to Ser294 and is positioned to interactwith SXN motif, residues Ser349 and Asn351 (Fig. 2A).

The bulky aminothiazole-containing group has extensive inter-actions with PaPBP3. The amide bond of the ceftazidime iswedged between the Thr487 carbonyl backbone and the Asn351side chain. The aminothiazole moiety is stabilized by hydrogenbonds with the Glu291 side chain and Arg489 carbonyl backboneand by hydrophobic interaction with Tyr409. Importantly, theTyr409 side chain blocking the aminothiazole binding site in theabsence of ligand is swung out toward the solvent-exposed surfaceto avoid a steric clash upon ceftazidime binding. The Tyr409,flanked by two highly flexible glycine residues, is also present inhigh-molecular mass PBPs of subclass B3 from other Gram-negative bacteria including N. gonorrhoeae PBP2 and E. coliPBP3, and similar conformational changes involving tyrosineresidues appear to occur upon the binding of aminothiazolemoieties (8, 14). In contrast, the aminothiazole moiety makesrelatively few direct interactions with β-lactamases and appar-ently adopts a different conformation in the binding site of lowmolecular mass PBPs, suggesting different recognition mechan-

Fig. 1. Overall structure of the high-molecular-weight class B PBP3 fromGram-negative bacteria. (A) Stereoview of the PaPBP3 complexed with MC-1.TheN-terminal domain is shown in orange and the C-terminal domain in cyan.The bound MC-1 is shown as spheres. (B) Chemical structures of antibioticsrelevant to this study. The leaving group of ceftazidime is shown in red.

Han et al. PNAS ∣ December 21, 2010 ∣ vol. 107 ∣ no. 51 ∣ 22003

CHEM

ISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1

isms for the aminothiazole moieties among different classes ofPBPs and β-lactamases (15, 16).

Unexpectedly, Tyr532 and Phe533 adopt a completely differentconformation upon ceftazidime binding, displacing the Tyr503side chain toward the active site. These changes result in a uniquearomatic wall composed of Tyr503, Tyr532, and Phe533 thataccommodates the bulky gem-dimethyl group appended from theoxime by a hydrophobic interaction. The aromatic wall is furtherstabilized by forming a bridge with Val333, describing the bound-ary of the pocket for the gem-dimethyl group. Notably, upon cef-tazidime binding, Val333 shifts by approximately 1.5 Å away fromthe active site, not only to interact with the gem-dimethyl groupbut also to allow van derWaals interaction with the six-membereddihydrothiazine ring in the active site (Fig. 2B). In addition, thecarboxylic acid group forms a salt bridge with the guanidiumgroup of Arg489. The key features of the aromatic wall, togetherwith Val333 and Arg489, are likely to play a role in the efficientinhibition of PaPBP3.

Conformational Rearrangement of Tyr532 and Phe533 upon Carbape-nem (Meropenem and Imipenem) Binding.Carbapenems such as imi-penem and meropenem possess a broad antibacterial spectrumand are used primarily to combat penicillin- and cephalosporin-resistant bacteria (17, 18). The electron density within the activesite revealed each carbapenem molecule covalently bound toSer294 (Fig. S2). Unlike the aminothiazole-containing ceftazi-dime complex, the Tyr409 side chain remains hydrogen-bondedto the Thr487 carbonyl backbone, where the Thr487 side chainis not close enough to make a direct hydrogen-bond interactionwith the C3 carboxylic acid group. Instead, the C3 carboxylic acidis anchored by a water-mediated hydrogen bond with Thr487.In both the imipenemandmeropenemcomplex structures, the for-mation of the complex was accompanied by substantial conforma-tional change in the regions of β3, β4 and the loop connecting β5andα11 at themouth of the active site. TheCα atomsofTyr532 andPhe533 in the imipenem complex were displaced 7.1 Å and 4.3 Å,respectively, toward the opening of the active site pocket com-pared to their positions in the apo structure (Fig. 2C). A simulta-neous reorientation of Tyr503 andArg489 results in formation of astacking interaction with Tyr409. As a result, the hydrophobic pyr-rolidine core of imipenem is completely buried by a hydrophobicpocket formed by residues Val333 and Tyr532.

Surprisingly, the additional methyl group of the pyrrolidinecore in meropenem triggers a 180° flip of the Tyr532 side chainand moves its side chain away from the active site to interact withthe Asn242 carbonyl backbone (Fig. 2D). Phe533 moves towardthe active site forming a tunnel-like hydrophobic pillar withVal333 to stabilize the methyl group of the pyrrolidine core. Thecarbamoyl pyrrolidinyl group of meropenem is further stabilizedby water-mediated hydrogen bonds with the Ile347 and Gly535backbone. In contrast, in the imipenem complex structure, theelectron density beyond the thioether sulfur atom is weak anddiscontinuous, suggesting that the C-2 side chain is not only flex-ible but also does not form strong interactions with PaPBP3.Furthermore, a competition assay using a labeled penicillin Gshows that imipenem competes 16-fold less efficiently thanmeropenem for inactivation of PBP3 by a fluorescent penicillinanalog (Table 1). Although the PaPBP3 crystal structures incomplex with carbapenems are products of inactivation reactions,the hydrophobic interactions involving two different residues,Tyr532 and Phe533, and additional water-mediated hydrogenbonds likely play an important role in carbapenem recognitionand positioning for acylation of Ser 294.

The Efficient Binding of Monobactam Aztreonam to PaPBP3. Themonobactam aztreonam possesses very high affinity for PBP3from a wide range of aerobic Gram-negative bacteria includingP. aeruginosa, but is inactive against Gram-positive bacteria dueto poor interaction with their essential PBPs (19). The structureactivity relationship of monobactams suggests that the 2-ami-nothiazole moiety of aztreonam contributes to the activity againstGram-negative bacteria, and anti-Pseudomonal activity is en-hanced by the addition of an iminopropyl carboxyl moiety. Thesulfonic acid group attached to the nitrogen of the β-lactam ringis responsible for β-lactam activation, and methylation at the C4position provides enhanced stability of the ring to β-lactamaseattack (20). The crystal structure of PaPBP3 in complex with az-treonam reveals that the bulky aminothiazole-containing moietyis in a similar position to that adopted in the ceftazidime complex,and the same induced-fit-conformational changes occur involvingresidues Tyr409, Arg489, and the aromatic wall composed ofTyr503, Tyr532, and Phe533 (Fig. 3A). β-lactam structural differ-ences ranging from an unfused ring in aztreonam, to a five-mem-bered fused ring in the carbapenems, or six-membered fused ringin ceftazidime, do not affect the hydrophobic interaction withVal333. The C4 methyl group of aztreonam does hinder rotationin the ring-open form and is stabilized by the hydrophobic patch

Fig. 2. Interaction of ceftazidime, imipenem, and meropenem in the activesite of the PaPBP3. (A) Active site in the ceftazidime-acyl PaPBP3 structure.Ceftazidime is shown in stick rendering with cyan carbons and hydrogenbonds as dashes. (B) Molecular surface of the PaPBP3 in the active site regionin complexwith ceftazidime (cyan). The apo PaPBP3 structure is shown in pink.(C) Active site of PaPBP3 bound to imipenem (orange). The loop connecting β5and α11 undergoing significant conformational changes is shown inmagenta.(D) Active site of PaPBP3 bound tomeropenem (navy). The loop connecting β5and α11 undergoing significant conformational changes is shown inmagenta.

22004 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1013092107 Han et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1

composed of Val333 and Phe533. In addition, the sulfonic acidgroup is tightly anchored by side chains of Lys484, Ser485, andThr487. Although the lactam cores of aztreonam and ceftazidimeare quite distinct, the bulky aminothiazole-containing group playsan important role in inducing conformational changes critical forefficient interaction with P. aeruginosa PBP3.

TheMonocarbam Siderophore ConjugateMC-1 and its in Vitro Activity.The monocarbams are monocyclic β-lactam antibacterial agentsthat have a carbonylaminosulfonyl activating group at the N-1position. Significant efforts have been devoted to themodificationof β-lactams by incorporating a siderophore moiety to promoteuptake by Gram-negative bacteria by exploiting the siderophoreactive transport system (21, 22).MC-1 is a siderophore-conjugatedmonocarbam with a carbonylaminosulfonyl activating group atN-1 position and a hydroxypyridone siderophore connected bya triazolone-derived linker (Fig. 1B). Compared to aztreonam,MC-1 was found to have exquisite potency against P. aeruginosaclinical isolates with a MIC90 of 0.5 μg∕mL (Table 1).

In the cocrystal structure of PaPBP3 with MC-1 at 1.64-Å re-solution, the hydrophobic pillar composed of Val333 and Phe533bisects the active site cleft opening with the iminopropyl carboxylmoiety on one side and the hydroxypyridone with triazolone linkeron the opposite side (Fig. 3 B and C). The bulky aminothiazole-containing moiety displaces the Tyr409 side chain and inducessimilar conformational changes as seen in the ceftazidime- andaztreonam-bound structures. The gem-dimethyl group is stabi-lized by a hydrophobic interaction with residue Val333 and thearomatic wall composed of Tyr503, Tyr532, and Phe533. The car-boxylic acid group forms salt bridges with Arg489. Notably, thetriazolone linker adopts two alternative conformations in a per-pendicular orientation, which allows the triazolone carbonylgroup to interact either with Gly535 backbone or with Leu346and Lys484 via a water molecule. The hydroxypyridone sidero-

phore is oriented toward the solvent-exposed surface and involvedin van der Waals interaction with Val333 and Phe533.

Unlike the sulfonic acid group directly attached to the N-1 po-sition of the β-lactam ring in aztreonam, the MC-1 has a longercarbonylaminosulfonyl species for activation. The additional car-bonylamino moiety pushes the sulfonyl group outside the activesite by 1.3 Å. As a result, the sulfonyl group in the MC-1 acyl-en-zyme structure forms two hydrogen bonds with backbone amidesof Gly534 and Gly535, effectively stabilizing a helix dipole at theN-terminal end of the α11 helix. Interestingly, the α11 helix con-tains a proline residue (Pro540) that kinks the N-terminal end ofhelix α11 toward the active site and helps the Phe533-containingloop forming one side of active site cleft. Mutation of the equiva-lent Pro540 to serine inN. gonorrhoeaePBP2 leads to a decrease inacylation rate despite no discernible effect on the structure of theα11 helix (8). Importantly, in the MC-1 acyl-enzyme, the addi-tional carbonylaminosulfonyl moiety forms an extensive hydro-gen-bond network with conserved residues. The Ser349 of theSXN motif forms a hydrogen bond with the carbonyl oxygen ofsulfonyl cabonylamino moiety, whereas Thr487, the fourth resi-due of the conserved KT/SGS/T motif, forms bidentate hydrogenbonds with the nitrogen atom and the other sulfonyl oxygen atom.

Solvent Rearrangement Effects and Second-Order Acylation RateConstants. A detailed molecular understanding of interactionsbetween proteins and ligands involves many factors (23). Barriersto ligand binding and dissociation can arise from solvent rearran-gement effects that can be rate determining when solvent transferto and from the binding site is critical (24). A consequence of thisfor covalent inhibitors is that the second-order acylation rate con-stants, which are indicative of the efficiency of the ligand to inac-tivate the target, may be predicted by the protein desolvationeffects that occur during ligand binding. This assumes that the en-ergetics of covalent bond formation and protein conformationaleffects do not significantly change the rank ordering of the ligands.

Table 1. In vitro characterization of β-lactams in PaPBP3 and hydration site analysis

CompoundSecond-order

acylation constantPaPBP3binding

Antibacterialactivity

Thermalstability

Stable water,ΔG < 0 kcal∕mol

Unstable water,ΔG > 0 kcal∕mol

Water displacementeffectiveness score

k2∕Kd (M−1 s−1) EC50 (uM) MIC90

(ug∕mL)Tm* (°C) ΔGi

s Nis hΔGi

si ΔGiu Ni

u hΔGiui ϵi logðk2∕KdÞ

M−1 s−1

Meropenem 63,760 0.157 >64 42 −3.563 10 −0.356 6.035 9 0.671 −0.531 4.805Imipenem 712 2.5 >64 40 −0.211 2 −0.105 6.473 11 0.588 −0.179 2.852Ceftazidime 24,707 0.17 >64 54 −2.597 6 −0.433 24.238 23 1.054 −0.411 4.393Aztreonam 4,396,400 0.086 >64 56 −3.935 4 −0.984 10.789 12 0.899 −1.094 6.643MC-1 45,108 0.109 0.5 58 −5.996 9 −0.666 30.395 24 1.266 −0.526 4.654

*Tm, thermal stability (apo PaPBP3 ¼ 46 °C); ΔGis ¼ ΣΔG, where (ΔG < 0 kcal∕mol) for ligand i; ΔGi

u ¼ ΣΔG, where (ΔG > 0 kcal∕mol) for ligand i; Nsi , number

of stable hydration sites affected by ligand i; Nui , number of unstable hydration sites affected by ligand i; hΔGi

si, unstable water displacement efficiency indexof ligand i; hΔGi

si, stable water displacement efficiency index of ligand i; ϵi ¼ hΔGisi∕hΔGi

ui, water displacement effectiveness score of ligand i.

Fig. 3. Aztreonam and the siderophore-conjugated monocarbam, MC-1 bound to PaPBP3. (A) Active site in aztreonam-acyl-PaPBP3 complex. Aztreonam isshown in stick rendering with magenta carbons. (B) Interaction ofMC-1 with PaPBP3.MC-1 is shown in stick rendering with two alternate conformations (greenand magenta). (C) Molecular surface of the PaPBP3 in the active site region in compex with MC-1 (two alternate conformation in green and magenta).

Han et al. PNAS ∣ December 21, 2010 ∣ vol. 107 ∣ no. 51 ∣ 22005

CHEM

ISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1

A computational approach, called WaterMap, computes thethermodynamic profile and free energies of hydration sites ina protein active site relative to bulk solvent and has been success-fully applied in other systems such as peptides binding to PDZdomains, kinase selectivity, and ranking of congeneric moleculesbinding to factor Xa (25–28). The approximation of free energy isbased exclusively on the displacement of water molecules withinthe site upon binding of the ligand and ignores other terms suchas protein–ligand van der Waals contacts, electrostatic interac-tions, internal strain (ligand and protein), protein conformationaleffects, and entropy changes. WaterMap energies also allow thecategorization and classification of the computed hydration sitesinto stable and unstable waters based on the thermodynamic pro-file of each computed water cluster. As ligand binding involvesthe exchange of protein–water for protein-ligand contacts, thedissection of the thermodynamics of hydration sites could provideinsights into ligand binding and thereby the efficiency of PBP3inactivation by siderophore-conjugated monocarbams (24).

Hydration sites were categorized into two different types ofwaters—stable (ΔG < 0 kcal∕mol) and unstable (ΔG > 0 kcal∕mol). Protein desolvation energies for each type of water site werecalculated for each ligand based on the respective crystal struc-tures (Table 1). A normalization process, in which the ligand-affected total free energy of hydration sites (ΔGi

u;s) was dividedby the number of ligand-affected hydration sites (Ni

u;s), standar-dizes this measure for any size effect of the ligand. This yields twoprotein desolvation efficiency indices: an unstable water displace-ment efficiency index hΔGi

ui and a stable water displacementefficiency index hΔGi

si. A composite water displacement effec-tiveness score (ϵi) is obtained by taking the ratio of hΔGi

si overhΔGi

ui. Both imipenem and meropenem binding do not involveinduced conformational changes of Tyr409 relative to the apo formand displace stable hydration sites (ΔG < 0) to varying degrees.Meropenemdisplaces 10 stable hydration sites, whereas imipenemdisplaces only 2 stable hydration sites. Both these compounds dis-place similar numbers of unstable hydration sites (ΔG > 0). Inaddition to the solvent rearrangement effects, imipenem andmeropenem induce significant protein conformational changes inTyr532 and Phe533 that were observed in the covalent-adduct for-mation (vide supra). In contrast, ceftazidime, MC-1, and aztreo-nam displace a significant number of unstable hydration sites(Fig. 4) and their unstable water displacement efficiency indexranges from 0.90 to 1.27 compared to 0.59 to 0.67 for imipenemand meropenem. MC-1 and ceftazidime effectively displace agreater number of unstable hydration sites than other β-lactams,whereas aztreonam and MC-1 effectively displace a greater num-ber of stable hydration sites. A plot between logðk2∕KdÞ and (ϵi)indicates a reasonable trend between second-order acylation con-stants and water displacement effectiveness scores (Table 1 andFig. S3).

For a simple one-step binding process where covalent bondformation is rapid, the rate-determining step is the formationof the encounter complex: a complicated process involving pro-tein–ligand recognition, protein conformational changes, anddesolvation effects. Depending on the compound and the bindingevent, the slowest, rate-determining step can be defined by oneof the many effects indicated above. It is interesting to note thatdespite the diversity of scaffolds involved in the core β-lactam ofthese five ligands and the various protein conformational changesobserved, the composite water displacement effectiveness score(ϵi) rank orders the five ligands according to their second-orderacylation constants (k2∕Kd). This composite score balances twotypes of desolvation energies—unstable water displacement effi-ciency index hΔGi

ui, which provides an indication of the boostthat is obtained by displacing loosely held waters that need notbe a specific recognition element, and the stable water displace-ment efficiency index hΔGi

si that captures specific pharmacopho-ric recognition elements by displacing tightly held waters. The

index (ϵi) appears to provide a measure of how effectively theligand binds in the initial interaction complex formed beforethe covalent bond formation, which is exclusively based on pro-tein desolvation energies.

Compounds that have more positive hΔGiui tend to stabilize

the covalent–adduct complex, as measured by thermal stabiliza-tion assay (Table 1), whereas compounds that do not stabilizethe covalent-adduct complex relative to the apo form have lowerhΔGi

ui. On the other hand, the stable waters are critical for ap-propriate recognition of key functionalities in the protein. Hencethe displacement of these waters with polar functional groups iscritical for both recognition and biological activity values. Com-pounds that have good submicromolar EC50 values all appear tobe replacing stable waters to a significant extent and have morenegative hΔGi

si. Despite its small size, aztreonam binds efficientlyby replacing most of the stable waters it is able to access with polarfunctionalities and has the most negative hΔGi

si and the mostpotent EC50 value. Significant thermal stabilization is also gainedfrom the covalent-adduct formation (in addition to the solventrearrangement effects), and the thermodynamics of covalent bondformation are not discussed here.

In conclusion, the unique conformation of PaPBP3 with β-lac-tam inhibitors represents a distinct high-molecular-weight class BPBP subfamily identifiable by the common features of the aro-matic wall formed by Tyr503, Tyr532, and Phe533 and the struc-tural flexibility of Tyr409 flanked by two glycine residues. Theguanidium group of Arg 489 shows a specific interaction withthe acid moiety of the inhibitors. This is the first structure ofa siderophore-conjugated triazolone-linked monocarbam com-plexed with PBP and provides molecular insights into the efficientinteraction of siderophore-conjugated monocarbams with theimportant therapeutic target PBP3. The analysis of energeticsand rearrangement of waters suggests that protein desolvationeffects contribute to PBP3 binding, and the water displacementeffectiveness score (ϵi) is indicative of the second-order acylationconstants. Our structural, biochemical, and computational studiesprovide a molecular basis for understanding the coupled activityand recognition specificity forP. aeruginosaPBP3 and for design ofinhibitors of this class of compounds.

Experimental Procedures.Complete details about the experimentalprocedures can be found in SI Text.

Ninety Percent of the Minimal Inhibitory Concentration (MIC90). Onehundred six recent P. aeruginosa clinical isolates were assayedusing broth microdilution, according to the Clinical and Labora-tory Standards Institute guidelines for antimicrobial susceptibilitytesting.

P. aeruginosa PBP3 Competition of Binding Assay. Two hundred na-nograms of purified PaPBP3 was assayed against various concen-trations of β-lactam and 0.65 μM Bocillin FL (at its apparent Km

Fig. 4. WaterMap results of P. aeruginosa PBP3 in complex with MC-1crystal structure. Hydration sites are depicted as spheres colored byΔG. Stablewaters,ΔG < 0, are blue and unstable waters, ΔG > 0, are red in color. MC-1 isshown in stick representation. Representative water sites are exemplified.

22006 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1013092107 Han et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1

value) at the same time (no preincubation), which allows forsimple competitive binding to the PaPBP3.

Rapid Quench Flow Experiment. Acylation rates of PaPBP3 and β-lactam antibiotics were determined using a RQF-3 Rapid QuenchFlow instrument (Kintek), as discussed in detail previously (29).

Protein Cloning, Expression, and Purification. A soluble form ofPaPBP3 construct (residues 50–579) appended with an N-term-inal His6 tag was cloned into pET28 (Novagen) and overex-pressed in E. coli BL21(Gold) in Novagen’s autoinductionmedium for 24 h at 25 °C. Purified PaPBP3 was concentratedto 10 mg∕mL for crystallization trials.

Crystallization.Crystals of apo-PaPBP3 were obtained with a reser-voir solution containing 30% PEG 4000, 0.2 MMgCl2, and 0.1 MTris pH 8.5. Each compound (ceftazidime, imipenem, merope-nem, and aztreonam) was soaked into the crystals. Cocrystalliza-tion of PaPBP3 with MC-1 was achieved with well solutionscontaining 0.1 M citrate; pH 6.2, 30% PEG 3350, and 0.3 MðMgNO3Þ2. Crystals were cryoprotected by dragging the crystalsthrough MiTeGen’s LV CryoOil™ (MiTeGen, LLC) and flashfrozen in liquid nitrogen.

Structure Determination. Diffraction data were collected fromflash-frozen crystals at 100 K at the Advanced Photon Sourceof the Argonne National Laboratory on a ADSC Quantum 210CCD detector. Data were processed using the HKL2000 suiteof software (30). Data collection statistics are summarized in

Table S1. The structure of apo-PaPBP3 was solved by molecularreplacement methods with the CCP4 version of PHASER (31),using Neisseria gonorrhoeae PBP2 [Protein Data Bank (PDB) IDcode 3EQU] as a search model. After molecular replacement,maximum likelihood-based refinement of the atomic positionand temperature factors were performed with REFMAC (32)and autoBUSTER (33), and the atomic model was built withthe program COOT (34). The refined PaPBP3-apo structurewas then used as a starting model for all other complexes. Thestereochemical quality of the final model was assessed by PRO-CHECK (35). Crystallographic statistics for the final models areshown in Table 1. Figures were prepared with PYMOL (36).

Thermal Stability Assay. The thermal shift assay was conducted inthe iCycler iQ Real Time Detection System (Bio-Rad), originallydesigned for PCR. The fluorescence intensity was measuredwith Ex∕Em: 490∕530 nm.

WaterMap Calculations of PBPs. WaterMap calculations were per-formed on the five ligand-bound crystal structures (aztreonam,ceftazidime, meropenem, imipenem, and MC-1) and the apoPaPBP3 crystal structure.

Coordinates. Coordinates for the PaPBP3 and inhibitor structureshave been deposited with the Protein Data Bank.

ACKNOWLEDGMENTS. We thank Lisa Mullins for determining MIC90 and JohnMueller, Mark Plummer, Mark Mitton-Fry, and Xiayang Qiu for insightfuldiscussions.

1. Giske CG, Monnet DL, Cars O, Carmeli Y (2008) Clinical and economic impact ofcommon multidrug-resistant gram-negative bacilli. Antimicrob Agents Chemother52:813–821.

2. Boucher HW, et al. (2009) Bad bugs, no drugs: No ESKAPE! An update from the Infec-tious Diseases Society of America. Clin Infect Dis 48:1–12.

3. Giamarellou H, Kanellakopoulou K (2008) Current therapies for pseudomonasaeruginosa. Crit Care Clin 24:261–278.

4. Siegel RE (2008) Emerging gram-negative antibiotic resistance: Daunting challenges,declining sensitivities, and dire consequences. Respir Care 53:471–479.

5. Liao X, Hancock RE (1997) Susceptibility to beta-lactam antibiotics of Pseudomonasaeruginosa overproducing penicillin-binding protein 3.Antimicrob Agents Chemother41:1158–1161.

6. Spratt BG, Cromie KD (1988) Penicillin-binding proteins of gram-negative bacteria. RevInfect Dis 10:699–711.

7. Denome SA, Elf PK, Henderson TA, Nelson DE, Young KD (1999) Escherichia colimutants lacking all possible combinations of eight penicillin binding proteins:Viability, characteristics, and implications for peptidoglycan synthesis. J Bacteriol181:3981–3993.

8. Powell AJ, Tomberg J, Deacon AM, Nicholas RA, Davies C (2009) Crystal structuresof penicillin-binding protein 2 from penicillin-susceptible and -resistant strains ofNeisseria gonorrhoeae reveal an unexpectedly subtle mechanism for antibiotic resis-tance. J Biol Chem 284:1202–1212.

9. Page MG, Dantier C, Desarbre E (2010) In vitro properties of BAL30072, a novelsiderophore sulfactam with activity against multiresistant gram-negative bacilli.Antimicrob Agents Chemother 54:2291–2302.

10. Zurenko GE, Truesdell SE, Yagi BH, Mourey RJ, Laborde AL (1990) In vitro antibacterialactivity and interactions with beta-lactamases and penicillin-binding proteins of thenew monocarbam antibiotic U-78608. Antimicrob Agents Chemother 34:884–888.

11. Goffin C, Ghuysen JM (1998) Multimodular penicillin-binding proteins: An enigmaticfamily of orthologs and paralogs. Microbiol Mol Biol Rev 62:1079–1093.

12. Sauvage E, Kerff F, Terrak M, Ayala JA, Charlier P (2008) The penicillin-bindingproteins: Structure and role in peptidoglycan biosynthesis. FEMS Microbiol Rev32:234–258.

13. Faraci WS, Pratt RF (1986) Interactions of cephalosporins with the StreptomycesR61 DD-transpeptidase/carboxypeptidase. Influence of the 3′-substituent. Biochem J238:309–312.

14. Wientjes FB, Nanninga N (1991) On the role of the high molecular weight penicillin-binding proteins in the cell cycle of Escherichia coli. Res Microbiol 142:333–344.

15. Powers RA, Caselli E, Focia PJ, Prati F, Shoichet BK (2001) Structures of ceftazidime andits transition-state analogue in complex with AmpC beta-lactamase: Implications forresistance mutations and inhibitor design. Biochemistry 40:9207–9214.

16. Kuzin AP, Liu H, Kelly JA, Knox JR (1995) Binding of cephalothin and cefotaxime toD-ala-D-ala-peptidase reveals a functional basis of a natural mutation in a low-affinitypenicillin-binding protein and in extended-spectrum beta-lactamases. Biochemistry34:9532–9540.

17. Lee VJ, Miller GH, Yagisawa M (1999) What’s new in the antibiotic pipeline. Curr OpinMicrobiol 2:475–482.

18. Quinn JP (1998) Clinical strategies for serious infection: A North American perspective.Diagn Micr Infec Dis 31:389–395.

19. Sykes RB, Bonner DP, Bush K, Georgopapadakou NH (1982) Azthreonam (SQ 26,776), asynthetic monobactam specifically active against aerobic Gram-negative bacteria.Antimicrob Agents Chemother 21:85–92.

20. Bonner DP, Sykes RB (1984) Structure activity relationships among the monobactams.J Antimicrob Chemother 14:313–327.

21. Miller MJ (1989) Syntheses and therapeutic potential of hydroxamic acid basedsiderophores and analogues. Chem Rev 89:1563–1579.

22. Mollmann U, Heinisch L, Bauernfeind A, Kohler T, Ankel-Fuchs D (2009) Siderophoresas drug delivery agents: Application of the “Trojan Horse” strategy. Biometals22:615–624.

23. Mobley DL, Dill KA (2009) Binding of small-molecule ligands to proteins: “What yousee” is not always “what you get”. Structure 17:489–498.

24. Pearlstein RA, et al. (2010) New hypotheses about the structure-function of proproteinconvertase subtilisin/kexin type 9: Analysis of the epidermal growth factor-like repeatA docking site using WaterMap. Proteins 78:2571–2586.

25. Young T, Abel R, Kim B, Berne BJ, Friesner RA (2007) Motifs for molecular recognitionexploiting hydrophobic enclosure in protein-ligand binding. Proc Natl Acad Sci USA104:808–813.

26. Abel R, Young T, Farid R, Berne BJ, Friesner RA (2008) Role of the active-site solvent inthe thermodynamics of factor Xa ligand binding. J Am Chem Soc 130:2817–2831.

27. Beuming T, Farid R, Sherman W (2009) High-energy water sites determine peptidebinding affinity and specificity of PDZ domains. Protein Sci 18:1609–1619.

28. Robinson DD, Sherman W, Farid R (2010) Understanding kinase selectivity throughenergetic analysis of binding site waters. ChemMedChem 5:618–627.

29. Lu WP, Kincaid E, Sun Y, Bauer MD (2001) Kinetics of beta-lactam interactionswith penicillin-susceptible and -resistant penicillin-binding protein 2x proteins fromStreptococcus pneumoniae. Involvement of acylation and deacylation in beta-lactamresistance. J Biol Chem 276:31494–31501.

30. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in oscilla-tion mode. Methods Enzymol 276:307–326.

31. McCoy AJ, et al. (2007) Phaser crystallographic software. J Appl Crystallogr 40:658–674.32. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular structures

by the maximum-likelihood method. Acta Crystallogr D 53:240–255.33. Blanc E, et al. (2004) Refinement of severely incomplete structures with maximum

likelihood in BUSTER-TNT. Acta Crystallogr D 60:2210–2221.34. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. Acta

Crystallogr D 60:2126–2132.35. Laskowski RA,Moss DS, Thornton JM (1993)Main-chain bond lengths and bond angles

in protein structures. J Mol Biol 231:1049–1067.36. Delano WL (2002) The PyMOL Molecular Graphics System (Delano Scientific LLC, Palo

Alto, CA) www.pymol.org.

Han et al. PNAS ∣ December 21, 2010 ∣ vol. 107 ∣ no. 51 ∣ 22007

CHEM

ISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1