Reactibodies generated by kinetic selection couplechemical reactivity with favorable protein dynamicsIvan Smirnova,b,1, Eugénie Carlettic,1, Inna Kurkovaa,1, Florian Nachond, Yvain Nicoletc, Vladimir A. Mitkeviche,Hélène Débatf,g, Bérangère Avalleh, Alexey A. Belogurov, Jr.a,i, Nikita Kuznetsovj, Andrey Reshetnyakk,Patrick Massond, Alexander G. Tonevitskyl, Natalia Ponomarenkoa, Alexander A. Makarove, Alain Fribouleth,Alfonso Tramontanom, and Alexander Gabibova,b,i,2

aInstitute of Bioorganic Chemistry Russian Academy of Sciences, Moscow 117997, Russia; bChemistry Department, Moscow State University, Moscow 119991,Russia; cInstitut de Biologie Structurale (Commissariat à l’Energie Atomique-Centre National de la Recherche Scientifique - Université Joseph Fourier), 38027Grenoble, France; dDepartement de Toxicologie, Institut de Recherche Biomédicale des Armées, 38702 La Tronche, France; eEngelhardt Institute of MolecularBiology, Russian Academy of Sciences, Moscow 119991, Russia; fUniversité Versailles Saint-Quentin, F-78035 Versailles, France; gInstitut de Génétique etMicrobiologie, Unité Mixte de Recherche 8621 Centre National de la Recherche Scientifique, Université Paris-Sud, F-91405 Orsay, France; hUniversité deTechnologie de Compiègne, Unité Mixte de Recherche 6022 Centre National de la Recherche Scientifique, 60205 Compiègne, France; iInstitute of Gene Biology,Moscow 117334, Russia; jInstitute of Chemical Biology and Fundamental Medicine, Siberian Branch of the Russian Academy of Sciences, Novosibirsk 630090,Russia; kDepartment of Pharmacology, Yale University School of Medicine, Yale University, New Haven, CT 06520-8066; lInstitute of General Pathology andPathophysiology, Russian Academy of Medical Science, 125315 Moscow, Russia; and mSchool of Medicine, University of California, Davis, CA 95616

Edited* by Michael Sela, Weizmann Institute of Science, Rehovot, Israel, and approved August 1, 2011 (received for review May 30, 2011)

Igs offer a versatile template for combinatorial and rational designapproaches to the de novo creation of catalytically active proteins.We have used a covalent capture selection strategy to identifybiocatalysts from within a human semisynthetic antibody variablefragment library that uses a nucleophilic mechanism. Specificphosphonylation at a single tyrosine within the variable light-chainframework was confirmed in a recombinant IgG construct. High-resolution crystallographic structures of unmodified and phospho-nylated Fabs display a 15-Å-deep two-chamber cavity at the inter-face of variable light (VL) and variable heavy (VH) fragments havinga nucleophilic tyrosine at the base of the site. The depth and struc-ture of the pocket are atypical of antibodies in general but can becompared qualitatively with the catalytic site of cholinesterases. Astructurally disordered heavy chain complementary determiningregion 3 loop, constituting a wall of the cleft, is stabilized aftercovalent modification by hydrogen bonding to the phosphonatetropinolmoiety. These features and presteady state kinetics analysisindicate that an induced fit mechanism operates in this reaction.Mutations of residues located in this stabilized loop do not interferewith direct contacts to the organophosphate ligand but can interro-gate second shell interactions, because the H3 loop has a conforma-tion adjusted for binding. Kinetic and thermodynamic parametersalong with computational docking support the active site model,including plasticity and simple catalytic components. Although rela-tively uncomplicated, this catalytic machinery displays both stereo-and chemical selectivity. The organophosphatepesticideparaoxon ishydrolyzed by covalent catalysis with rate-limiting dephosphoryla-tion. This reactibody is, therefore, a kinetically selected protein tem-plate that has enzyme-like catalytic attributes.

catalytic antibodies | crystal structure | insecticide | organophosphatehydrolysis

Enzymes achieve extraordinary catalytic efficiencies by co-ordinating binding forces and reactive residues within their ac-

tive sites to alter themechanism and energy landscape of a chemicalreaction (1, 2). This functional complexity poses a challenge forapproaches to create artificial enzyme-like catalysts. The interplaybetween binding and catalysis can be explored through directedevolution strategies (3). Simple enzymatic activities have been de-veloped by adaptation of a protein template for binding to stableanalogs of substrates or transition states (TSA), which was exem-plified by the generation of a broad array of catalytic antibodies (4–6), and someof these anti-TSAantibodies use a nucleophilic residuein their catalytic mechanism (7). This approach has only excep-tionally been directed to recruit residues that participate in covalentcatalysis that can be considered essential for multistep catalytic ef-ficiency (8). However, aldolase antibodies obtained by reactive im-

munization display among the most efficient de novo generatedprotein catalytic activity yet reported, and structural studies of theseantibodies have shown how their binding mechanism allows fordiscontinuous evolution of a promiscuous catalytic site (9, 10).Selection by reversible covalent binding is compatible with

conventional affinity maturation, where the protein environmentprovides thermodynamic stabilization of the covalent adduct. Al-ternatively, irreversible covalent binding allows for kinetic selec-tion, favoring sites that accelerate bond formation by transitionstate stabilization. In a classical example, serine hydrolases catalyzephosphorylation of their active site as a result of the specificallyenhanced reactivity of the nucleophilic serine-195 (chymotrypsin-ogen numbering system) (11). We have sought to use such en-hanced reactivity for the generation of a reactibody. It is, thus,evident that innate reactivity can be deployed to capture morehighly evolved biocatalysts that conserve an essential residue forcovalent catalysis (12). Mechanism-based kinetic selection may,thus, be differentiated from suicide substrate selection or similarstrategies that have successfully used covalent capture (13). Wepreviously applied this approach to selection of a diversified rep-ertoire of human Ig variable light (VL) and variable heavy (VH)single-chain fragments (scFv) displayed on phage particles (14).Using a biotinylated phosphonate ester (1a) (Fig. 1) for solutionphase modification, we identified a set of nucleophilic scFvencompassing sites predisposed for covalent catalysis. Sequenceanalysis of the reactive clones revealed preferred pairing of λ-VLand VH chains with conserved heavy chain complementary de-termining region 3 (CDR-H3) sequences. Two structural motifswere seen using distinct tyrosine residues in VL as nucleophile. Ofthese clones, seven of eight clones had a tyrosine (Y-L33) residuewithin CDR1 as a putative nucleophile. However, the most nu-cleophilically active clone, A.17, used a tyrosine residue (Y-L37) ina conserved framework (FR2), a clear expansion of the conven-tional concept of a binding site. The term reactibody is proposed

Author contributions: I.S., A.A.B., N.P., A.F., A.T., and A.G. designed research; I.S., I.K., V.A.M.,N.K., and A.R. performed research; E.C., F.N., Y.N., V.A.M., H.D., B.A., N.K., and A.A.M.contributed new reagents/analytic tools; I.S., E.C., I.K., F.N., Y.N., V.A.M., H.D., B.A., A.A.B.,N.K., A.R., P.M., A.G.T., N.P., A.A.M., A.F., A.T., and A.G. analyzed data; and I.S., I.K., A.A.B.,P.M., A.F., A.T., and A.G. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Freely available online through the PNAS open access option.

Data deposition: The crystallography, atomic coordinates, and structure factors have beendeposited in the Protein Data Bank, www.pdb.org (PDB ID codes 2XZA and 2XZC).1I.S., E.C., and I.K. contributed equally to this work.2To whom correspondence should be addressed. E-mail: gabibov@mx.ibch.ru.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1108460108/-/DCSupplemental.

15954e15959 | PNAS | September 20, 2011 | vol. 108 | no. 38 www.pnas.org/cgi/doi/10.1073/pnas.1108460108

Dow

nloa

ded

by g

uest

on

June

25,

202

0

here to denote a protein template selected for chemical reactivityrather than ground state binding.

Results and DiscussionTo facilitate efforts to elucidate structure–function correlation ofthe selected reactibodies, the VL and VH elements of A.17 werereconstructed for expression as full-length IgG inChineseHamsterOvary (CHO) cells (Fig. S1). The apparent nucleophilic reactivityof the scFvA.17was retained in the whole IgG construct, which wasdetermined using kinetic studies with phosphonate substrate 1b(Fig. 1). Analysis of the kinetic data for reaction between IgG A.17and 1b, usingDYNAFIT software (15), is compatible with a kineticmodel proceeding through a single noncovalent intermediate be-fore the rate-determining covalent step (1). Calculation of the

dissociation constant (KD, step 1) and rate constant for A.17phosphonylation (k2, step 2) revealed a catalytic efficiency similarto the efficiency of serine proteases reacting with phosphonateester inactivators (16).Although hydrolytic turnover was not observed, the reaction

scheme can be extended to include a step for hydrolysis of theintermediate to regenerate the nucleophilic tyrosine (step 3), andstructural studies can be implemented to provide a rational de-sign basis to implement this step (1):

Re Ab−Tyr−OHþOPX�����! �����

1:KD Re Ab−Tyr− −OPX����!2:k2 Re Ab−Tyr−OPþX −

Re Ab−Tyr−OPþ X −����!3:k3ðNu− Þ

Re Ab−Tyr−OH þNu− ; [1]

where Re Ab = reactibody, X = leaving group, and Nu = water/nucleophile.To clarify the molecular mechanism of the machinery of the

reactibody, crystal structures of the native A.17 Fab (Protein DataBank accession code 2XZA) and its product modified by reactionwith 1b (Protein Data Bank accession code 2XZC) have beensolved at 1.5- and 1.36-Å resolution, respectively (Table S1). Thesestructures reveal a deep cavity at the interface between VL (threeβ-sheets) and VH (four β-sheets) with several striking features (Fig.2 and Movie S1). Most notably, the antigen-binding site seems tobe subdivided into upper and lower chambers (Fig. 2 A–C), eachchamber formed from VH and VL residues. The upper chamber(Fig. 2A), defined primarily by five tyrosines (Y-H33, Y-H53, Y-H59, Y-L33, and Y-L50) plus tryptophan W-L92 (Fig. 2D), isanalogous to a specific tyrosine cluster typical of antibodies inter-actingwithDNAor other phosphoryl esters (17, 18). Anopening atW-L92, Y-H34, and N-H105 leads to a lower chamber consistingprimarily of framework residues. ReactiveY-L37 forms the base ofthe chamber (Fig. 2B) with its side chain packed against F-L100,W-H48, and W-H109, which orientates the hydroxyl group ofY-L37 into the cavity at a depth of 15 Å from the surface. Withinthis subsite, main chain atoms of G-H90/T-H91 and side chains ofS-L35, W-L92, P-L98, and F-L100 form a niche for the phenylgroup of the phosphonate ester, whereas H-N105 provides an Hbond to the pseudotropine nitrogen in the OP-A.17 structure (Fig.

4 65

2 31

N

OPPhO

O

NO2

CH3

R

OP

O O

FSO OF

NH2

O P O

O

SN

OPO

O

O

NO2

OPHO

O

O

NO2

N

Fig. 1. Chemical structures of compounds used in the study: p-nitrophenyl 8-methyl-8-azabicyclo[3.2.1]octyl phenylphosphonate (1b; R = H) and its bio-tinylated derivative (1a; R = biotinyl); diisopropyl fluorophosphate (DFP; 2);4-(2-aminoethyl)benzenesulfonyl fluoride (AEBSF; 3); 2-diethoxyphosphor-ylthioethyl-trimethylammonium iodide (echothiophate; 4); O,O-diethyl O-(4-nitrophenyl) phosphate (paraoxon; 5);O-(4-nitrophenylphosphoryl) choline (6).

Y-H59

Y-H53

Y-H33

Y-H34

W-H48S-H51

W-H109

W-L92

F-L100

Y-L37

N-H105

A-H107

Y-H59

Y-H53

Y-H33

Y-H34

W-H48

S-H51

W-H109

W-L92

F-L100

Y-L37

A B C

D E

Y-H33

Y-H53

Y-H59Y-L33

Y-L50

Y-L92

Y-H34

N-H105

Y-L37

3.2Å

2.5Å

2.8Å

Fig. 2. Structural analysis of reactibodyFabA.17. Crystallographic snapshots of thetwo-chamber active center. α-Chain tracerenderings on the upper (A) and lower (B)chambers and merged (C) of the substratebinding site of theA.17 active site shown inspace-fillingmode. The light chain is shownin pale blue, and the heavy chain is shownin green for A–C. A top view showing aro-matic residues forming the walls of theupper cavity and lower chamber centeredon reactive side chain of Y-L37 in the un-modified (D) and phosphonate-modified(E) structures. The light chain is shown ingray, the heavy chain is in blue, the VH loopis in magenta, and the OP molecule is ingreen. Key residues are represented assticks, with nitrogen atomsmarked by deepblue, oxygen atoms marked by red, andphosphorus marked by orange. A boundchloride ion, indicated by the green sphere,is constant to modification (Fig. S3A).

Smirnov et al. PNAS | September 20, 2011 | vol. 108 | no. 38 | 15955

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

0

2E). The dimensions and topology of the site are atypical of anti-bodies in general, including catalytic antibodies, which are illus-trated in Fig. 3 (19–21). The reactive nucleophile is farther fromthe protein surface than the corresponding group in an aldolaseantibody (15 vs. 13 Å). The depth and structure of the pocket maybe compared qualitatively with the catalytic sites of cholinesterases(Fig. 3) (22–24).Differences in the structure of native and 1b-modified reacti-

body suggest that considerable conformational rearrangementaccompanies ligand migration into the site (Fig. 4A). Basically,there are two primarymodels of themechanism of conformationalchange because of antibody–ligand interactions: the existence ofpreequilibrium conformers and an induced fit mechanism (25–27). Conformational rearrangements of reactibody A.17 have beenexamined by both structural methods and kinetic approaches toobtain a model that best fits one of these modes.Structural analyses revealed significant differences between

crystal structures of unmodified and phosphonylated antibodies.Residues L99-G110 of HCDR3 seem to define a conformationallymobile element that spans the upper and lower chambers, with key

residues A-H107 andW-H109 oriented into the cavity (Fig. 2E). Inthe unmodified structure, this solvent-exposed loop is highly dis-ordered, with a B factor of 40–50 Å2 compared with an overall Bfactor of 19 Å2. By contrast, the electron density map at theseresidues in OP-A.17 is well-defined, with B factors ranging from 15to 25 Å2 (Fig. S2). Additional displacements are required atH-H104 (3.05 Å), N-H105 (3.63 Å), A-H107 (0.70 Å), and W-H109 (0.86 Å) to accommodate the ligand and permit strongH bonding (2.79 Å) between the main-chain oxygen of N-H105and the pseudotropine nitrogen (Figs. 2E and 4A). Significantdisplacements in the L-chain, including a 20° χ2 angle rotationand 0.4-Å shift at W-L92 and a 10° rotation at Cα of F-L100,enable nesting of the phenyl ring of 1b with T-stacking inter-actions with the ligand. Thesemotions may be concerted with a 5°rotation of the Y-L37 side chain. These observed conformationalrearrangements support the existence of an induced fit-likemechanism during the ligand–antibody interaction. A cloud ofelectron density just above the face opposite to the Y-L37-OPbond was modeled as a cluster of mobile water molecules. Y-H34participates in an H-bonding network stabilizing this cluster (Fig.4B). The phosphonyl oxygen is strongly H-bonded to water w614(2.55 Å), suggesting polarization of the P–O bond to enhanceelectrophilicity (Fig. 2E). In unmodified A.17, a network of watermolecules (w689, w688, and w274) bridges the hydroxyl group ofS-L35 to N-H105 (Fig. S4). Conversely, transfer of the OP sub-strate into the lower compartment must break this network anddisplace these water molecules.In order to further assess interactions in the protein environ-

ment, thermodynamic parameters for the reaction of A.17 andselected mutants with 1b were measured by isothermal titrationcalorimetry (ITC) (Fig. S5A). The data were fitted to theoreticalbinding curves assuming a single ligand site. In case of A.17 bothentropy and enthalpy of reaction were favorable, with ΔH de-creasing threefold as temperature increased from 10 °C to 37 °C(Table 1). The observed enthalpy changes and apparent KD (KD

†)were in line with ITC-estimated parameters for antibodies bindingto their haptens (28). Furthermore, a theoreticalΔH of−2.56 kcal/mol for computational docking of 1b from the surface to the lowerchamber is in good correspondence with the experimental valueof −2.78 kcal/mol (Fig. S6 and Movie S2).Mutants Y-L33F and Y-L37F of A.17 IgG probe two possible

nucleophilic residues identified in scFv reactibodies (14). Consis-tent with previous studies and the X-ray structure, only Y-L37Fresults in loss of reactivity. TheΔH andKD

† ofA.17 andΔH andKDof the inactive mutant A.17 Y-L37F interacting with 1b (Fig. S5B)were nearly identical and within a factor of two of the respectivefactors for the BChE reaction (Table 1 and Fig. S5C). A very smallenthalpy changemight be expected for the ester exchange reactionwhere a P–OAr bond is broken and another formed to a tyrosineresidue (or to a serine residue in the case ofBChE).A similar effecthas been described in calorimetric studies of horse serum BChEinhibition by 5 (29). The covalent reaction ofA.17must also reflectfree energies caused by noncovalent interactions stabilizing theadduct and free energies caused by release of the leaving group.The latter energy change seemed to be negligible and was esti-mated in reference experiments of p-nitrophenol binding to nativeor chemically modified A.17 (Fig. S5D). These observations sug-gest that the thermodynamic changes are sensitive primarily tononcovalent interactions. Accordingly, changes in heat capacitybecause of binding, calculated from the temperature dependenceof enthalpy changes (Fig. S5E), were used to estimate the change insolvent accessible area (30) using the following empirical formula(Eq. 2):

ΔCp¼ −70cal

mol$K¼ 0:27ΔAar þ 0:4ΔAnonar; [2]

where ΔAar and ΔAnonar are the protected areas in angstromssquared because of formed by aromatic and nonaromatic residues,respectively (31). Because covalently bound OP is effectively bur-ied in the cavity, these changes can be attributed to conformational

TEPC-15

6Å

BChE

AChE

49G7

13Å

18Å

12Å

A.17

15Å

16Å

33F12

Fig. 3. Comparison of active site cavities of natural and de novo createdbiocatalysts. Chemically selected reactibody FabA.17 has a deep substratebinding niche. Cross-section views of the active center of esterolytic anti-bodies 49G7, TEPC15, aldolase antibody 33F12, choline esterases AChE andBChE, and antibody A.17 complexed with their ligands. In each case, thedistance measured is the height of a pyramid with a triangle base con-structed on the three residues nearest to the entrance of the active site, andapexes are the residue nearest to the ligand.

w614

w668

w646

w588

w648w363

Y-L37

Y-H34

W-L92

F-L100 Y-L37

W-H109

N-H105

H-H104

A-H107

A B

2.79Å

2.55Å

Fig. 4. Superposition of active sites of native (green) and OP-modified (blue)FabA.17 (A), illustrating the conformational changes of Y-L37, W- L92, F-L100,and residues of the VH loop (H-H104, N-H105, A-H107, and W-H109). Phos-phonate moiety is in red. (B) Representative view of the active site of 1b-modified A.17 Fab showing a strong H bond between a nonbridging phos-phonyl oxygen andwatermolecule w614 (2.55 Å). H-Y34 stabilizes a cluster ofsix water molecules by participating in their H-bond network. The light chainis in gray, the heavy chain is in blue, and theOPmolecule is in green. Hydrogenbonds are represented by gray dashes. A bound chloride ion, indicated by thegreen sphere, is independent of modification (Fig. S4).

15956 | www.pnas.org/cgi/doi/10.1073/pnas.1108460108 Smirnov et al.

Dow

nloa

ded

by g

uest

on

June

25,

202

0

changes at the protein surface. The ΔSAA of −175 to −260 Å2

calculated from ITC is comparable with ΔSAA of −240 Å2 calcu-lated from the crystal structures. In contrast to A.17WT, the ΔH ofthe Y-L37F interaction with 1b is independent of temperature,suggesting slight conformational changes in case of incubation ofthe Y-L37F antibody with 1b (Fig. S5E).Mutants N-H105A and H-H104A were also generated to in-

vestigate the role of the conformationally flexible CDR-H3 loop onreactivity. These mutations do not interfere with direct contacts tothe OP ligand but can interrogate second shell interactions as theCDR-H3 loop adjusts for binding. ITC data show similar KD

† butdiffering contribution of ΔH and ΔS for the reactions of 1b withA.17WT and A.17 H-H104A, where the latter reaction is en-thalpically more favorable (Table 1 and Fig. S5F). The effect ofboth mutations on the kinetically determined KD is also minimal,whereas the rate constant k2 decreases 2.4-fold in H-H104A andincreases 1.3-fold in N-H105A (Table 2). Thus, mutations withinthe H3 loop influence the pseudofirst order reaction rate withoutsignificantly altering the binding step 2 [1]. These results support anactive site model where H3 loop dynamics allow penetration ofreactant from outer to inner chambers. The H-H104 side chaintoggles between intrachainH bonds toD-H106 in unmodifiedA.17and to T-H100 in A.17-OP (Fig. S7). The H-H104A substitutionmay increase H3 conformational freedom but also can impair in-duced fit by destabilizing a conformation favored in the product.The N-H105 side chain participates in H bonding to the waterbridge with Y-L37 in the unmodified structure (Fig. S4). N-H105Aweakens this lattice, which could facilitate displacements for in-sertion of substrate into the reactive center. Based on theseobservations and data from crystallographic analysis, we suggestthat covalent modification of reactibody A.17 by 1b can be de-scribed by an induced fit model. To clarify this model, we carriedout presteady state kinetic investigations of fluorescence changingduring the phosphonylation reaction. According to steady statekinetics, theminimal kinetic scheme of the reaction includes stages

1 and 2 of equation 1 (Fig. 5A). The conformational fluctuations ofAb were monitored through intrinsic tryptophan fluorescencechanging during binding of phosphonate molecule to A.17. Weobserved a single increase in tryptophan fluorescence in a 100-mstime range (Fig. 5B Left). The direct dependence of kobs on sub-strate concentration suggests the prevalence of an induced fitmechanism in the reaction of A.17 with phosphonate 1b (Fig. 5BRight) (25). Overall, our balanced view is that the evidence favorsan induced fit-like model, whereas we can accept that, over thewide range of conditions that we have used, a significant fraction ofthe flux can occur through both pathways, resulting in a mixedmechanism for this reaction (32).The efficiency of A.17 reacting with 1b is more than an order of

magnitude greater than the corresponding BChE reaction (Table2), and it compares favorably with typical rates of serine proteasemodification by phosphonate 1b (16, 33, 34). This kinetic advan-tage is a clear demonstration that the complementary matching ofthe shape and chemical reactivity of substrate 1b with the selectedreactibody exceeds the performance of an enzyme such as BChEwith a classical phosphorylating agent. Enzyme selectivity for thephosphonate structure may also distort these comparisons (33–36). Reactive phosphoesters diisopropyl fluorophosphate (DFP)(2), paraoxon (5), and 4-(2-aminoethyl)benzenesulfonyl fluoride(AEBSF) (3) inhibited A.17 reaction with 1a, whereas echothio-phate (4) had no effect at concentrations up to 10mM (Fig. 6A andFigs. S3B and S8A). These molecules may be excluded from thereactive center while binding with varying affinities to the outersite. Stereoselective SN2 reaction with theRP enantiomer of 1bwasdeduced from the configuration at phosphorus in the A.17-OPadduct. Rate constants could be higher by a factor of two if the SPenantiomer does not react or even more if it is an inhibitor. Overa longer time span, hydrolysis of paraoxon 5 was detected, sug-gesting that a smallerOPmolecule can access the lower chamber ina productive orientation. Moreover, p-nitrophenol product wasreleased in molar excess to active sites (Fig. S8B), and the hydro-lysis rate was linearly dependent on A.17 concentration (Fig. S3C).The reaction rate also increased linearly with hydroxylamineconcentration (Fig. 6B), whereas mutant Y-L37F had no activity(Fig. S3D). These observations strongly support a reaction mech-anism that proceeds through a phosphotyrosine covalent inter-mediate at Y-L37. Elementary rate constants of the reaction (k2 =1.1± 0.1× 10−1 min−1, k3 = 1.6± 0.2× 10−2 min−1, and k4 = 1.26±0.09× 10−3mM−1min−1) were estimatedwhere dephosphorylationis rate limiting (Fig. 6B). Interaction between A.17 and paraoxonleads to the accumulation of the covalent intermediate accordingto the kinetic data. The existence of the covalent intermediate,supporting the covalent catalysis mechanism for paraoxon hydro-lysis by the A.17 reactibody, was also proved by MS analysis show-ing the existence of phosphorylated antibody (Fig. S8C).By contrast, the A.17Y-L37-OP ester obtained from reaction

with 1b was stable to hydrolysis under both native and denaturingconditions, whereas hydroxylamine (1–100 mM) had no influenceon the observed rate (Fig. S3E). Monoisonitrosoacetone, a classi-cal cholinesterase reactivator, also failed to regenerate A.17 re-activity. Thus, notwithstanding their chemical similarity, the twoA.17 phosphoesters show quite different P–O bond stability, sug-gesting that the active site assists hydrolysis of the intermediategenerated from 5. Although the mechanism remains speculativeand the rate of the phosphate release is slow, these data show thata pathway is available for catalytic turnover, which is described byequation 1.The enhanced reactivity of amino acid residues used in covalent

catalysis is an inherent function of the structured environmentwithin the enzyme active site. Reactive selection shows howarchitectures that support efficient chemistry can arise from rela-tively minor alterations of the protein interior. Studies of aldolaseantibodies obtained by reactive immunization suggested the re-active nucleophile could emerge from a single mutation releasinga lysine side chain constrained in a hydrophobic environment bya salt bridge or H bond (10, 21). The A.17 reactibody reveals analternative solution to reorganizing available structures for re-

Table 1. Thermodynamic parameters of A.17 WT, A.17 Y-L37F,A.17 H-H104A, and BChE binding to 1b determined by isothermaltitration calorimetry

Abzyme/enzyme

Temp(°C)

KD(KD†)

(μM)ΔH

(kcal/mol)TΔS

(kcal/mol)ΔG

(kcal/mol)

A.17 WT 10 2.9 −0.91 6.26 −7.17A.17 WT 15 3.5 −1.18 6.02 −7.20A.17 WT 20 2.8 −1.45 6.00 −7.45A.17 WT 25 2.3 −1.89 5.78 −7.67A.17 WT 37 1.7 −2.78 5.43 −8.21A.17 Y-L37F 10 1.4 −1.49 6.08 −7.57A.17 Y-L37F 15 3.2 −1.37 5.88 −7.25A.17 Y-L37F 25 1.0 −1.50 6.65 −8.15A.17 Y-L37F 37 0.72 −1.35 7.28 −8.63A.17 H-H104A 25 3.8 −3.6 3.72 −7.32BChE 15 7.9 −2.40 4.29 −6.69BChE 37 6.1 −4.35 2.93 −7.28

Table 2. Titration of the active sites of A.17, its mutants, andBChE with 1b (Kitz–Wilson)

Reactibody/enzyme k2 (min−1) KD (μM)k2/KD

(min−1 M−1 × 10−4)

A.17 WT 0.25 ± 0.02 6.2 ± 0.4 4.0 ± 0.6A.17 Y-L37F Inactive Inactive InactiveA.17 H-H104A 0.14 ± 0.01 8.5 ± 2.5 1.6 ± 0.6A.17 N-H105A 0.32 ± 0.01 7.7 ± 1.8 4.2 ± 1.1BChE* 0.066 ± 0.006 24 ± 6 0.3 ± 0.1

*The data obtained for BChE are consistent with previously reported data (34).

Smirnov et al. PNAS | September 20, 2011 | vol. 108 | no. 38 | 15957

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

0

activity. In antibody structures,Y-L37 is invariably buried at theVL-VH interface, where it packs against W-H109 (103) and adjacentresidues at the base of CDR-H3 (e.g., Cr6361, Protein Data Bank3GBM, 85%VL homology; E51, ProteinDataBank 1RZF, 84%VLhomology). By contrast, a long and flexible H3 loop in A.17 allowssufficient displacement at its base to open a chamber directly abovethe Y-L37 side chain. It is worth noting that H3 sequences weresynthetically diversified in the scFv library from which A.17 wasderived (37). It is conceivable that this diversity encompassessequences that are rare or prohibited in a natural repertoire.The disordered H3 loop in unmodified A.17 seems to act as

a flexible wall of the cavity to facilitate migration of ligand into thelower chamber. Moreover, compression and stabilization of thephosphonylated protein, evident from both the structure analysisand ITC studies, were indicative of a correlation between reactivityand protein dynamics, and they suggest that structural plasticity isa selected feature. Conformational changes in antibody bindingare not uncommon and may be seen with protein antigens whereflexibility can optimizemultiple interactions over a large antibody–antigen interface. However, antibodies elicited against haptens orother small ligands show evidence of diminishing V-region plas-ticity, because somatic mutations serve to anchor topologies suit-able for ligand complementarity (38, 39). Plasticity encoded in theV germline is thought to amplify combinatorial diversity, becausevarious shapes could be imprinted on each template. Accordingly,affinity maturation favors more rigid protein structures and maypreclude dynamic features in enzymatic function. By contrast, ki-

netic selection may exploit protein dynamics, exposing reactiveresidues in the protein interior. It is remarkable that motions deepwithin the pocket also enhance substrate complementarity. Al-though it is possible that a favorable enthalpy of chemical reactiondrives these displacements, the similar free energy of binding toa chemically inert Y-L37F mutant suggested that noncovalentinteractions may promote conformational changes for a produc-tive Michaelian complex. Collectively, our observations provide acompelling case for coupling between dynamics of the cavity andreactivity at the tyrosine side chain.Details of the structure capable of revealing interactions with a

putative trigonal bipyramidal transition state for ester exchangeand a trajectory for the leaving group are not evident from thestructures analyzed herein. However, small adjustments in contactresidues are clearly capable of accommodating such a transitionstate. The water cluster opposite the new P–Obond in the complexmay also play a role, and one water molecule is poised to hydrolyzethe phosphotyrosine ester to achieve turnover. Although there isno evidence that the site can activate this water for such nucleo-philic attack, a mechanism must be available to accomplish theobserved hydrolysis of the intermediate formed by reaction withparaoxon 5. Possibly, the site is locked in the case of reaction with1b, whereas the interaction with 5 allows dynamic displacementsfor the second-step reaction. Additional studies to clarify thisquestion may uncover possible means of improving the observedcatalytic efficiency.We set out to show that kinetic selection for covalent reactivity

can identify protein sites correctly configured for development ofartificial biocatalysts. Our thermodynamic and structural analysesunexpectedly revealed an associated feature with broader impli-cations. Recently, the question of how enzymes harness proteindynamics for catalytic efficiency has been revisited throughstructural and computational approaches (1, 40). Coupled con-formational substates optimized to admit or release substrates or

A

0 200 400 600 800 1,000 1,200 1,400 1,600 1,800

0

2

4

6

8

10

12

Time, s

Experimental trace

Fitted by model 1

Fitted by model 2

-0.030

0.000

0.030

Res

idua

ls

-0.030

0.000

0.030

Res

idua

ls

0 400 800 1,200 1,600

Ab + OPX KD

Ab--OPX Ab-OP k2

Model 2

0 400 800 1,200 1,600

Ab + OPX Ab-OP k2

Model 1

C[p

-nitr

ophe

nol],

0 10 20 30 40 5010

15

20

25

30

35

40

45

50

55

k obs

[OPX], µM0.01 0.1

Time, s

B Ab + OPX (Ab--OPX)’ (Ab--OPX)’’

1.8

2.0

2.2

Trp

fluor

esce

nce,

a.u

.

OPX, 1.0

3.0

10.0

20.0

40.0

50.0

KD k2

k-2

KD = 17.4±3.3 µMk2 = 54.9±2.7 s-1

k-2 = 12.1±1.3 s-1

µM

µM

Fig. 5. Discrimination of the modification mechanism between A.17 reac-tibody and phosphonate 1b. Experimental and fitted kinetic curves of in-teraction of A.17 with 1b. The quality of fit of kinetic models to theexperimental data was assessed by monitoring residuals against time fordifferent scheme fits (A). Presteady state kinetics curves of the intrinsic Trpfluorescence changing of A.17 depending on concentration of phosphonate1b (B Left). Dependence of kobs of the 1b interaction with reactibody A.17 onsubstrate concentration (B Right). A Inset shows the reaction mechanismused for data fitting and elementary constant calculation.

A.17+1a

a-Fc-HRP

Strept-HRP

DFP (2)

AEBSF (3)

Echothiophate (4

)

Paraoxon (5

)

A

k cat

. 102 , m

in-1

B

0 5 10 15 20 25

Ab + OPX KD

Ab--OPX Ab-OP

P1

Ab + P2

Ab + P3

k2 k3

k4 [NH2 OH]

k4

k3

4.0

3.0

2.0

1.0

5.0

0

Fig. 6. Description of multistep covalent catalysis by A.17 reactibody. Co-valent reactivity of A.17 and its mutants with biotinylated phosphonate 1a.Antibodies were incubated for 1 h at 37 °Cwith PBS buffer alone or 1mMeachof 2 (DFP) and 3 (AEBSF) and for 16 h with PBS buffer alone or 1 mM each 4(echothiophate) and 5 (paraoxon). All samples were then incubated for 1 h at37 °C with 100 μM of 1a and analyzed by Western blot (A). Concentrations ofantibodies were normalized as confirmed by comparable staining of heavychains on the same blots (A Upper). Three-step reaction of A.17 antibody with5 in the presence of NH2OH (B).

15958 | www.pnas.org/cgi/doi/10.1073/pnas.1108460108 Smirnov et al.

Dow

nloa

ded

by g

uest

on

June

25,

202

0

organize the active site for catalysis have been described (41).Reactibodies can present unique cavities that facilitate simplechemical processes and provide a useful model for how proteinscan acquire these adaptations. Strategies to deploy antibodies asenzymes have been criticized on the grounds that an Ig templatecannot deploy residues for covalent catalysis and that it lacks theflexibility required for dynamic mechanisms. Reactive immuni-zation and kinetic selection have provided evidence to dispel theseconcerns. At the same time, our reactibody template has provedadaptable in ways that may apply also to other protein scaffolds.

MethodsCrystallization of FabA.17 and OP-FabA.17. The purified A.17 Fab fragment (9mg/mL) was crystallized using the hanging drop method. Diffraction datawere collected at the European Synchrotron Radiation Facility on ID14-eh4beam line using λ = 0.979-Å wavelength with ADSC Quantum 4 detector.

ITC. The thermodynamic parameters of A.17WT, A.17 mutants, or BChE fromhuman plasma reacting with 1b or p-nitrophenol were measured on aMicroCal iTC200 instrument (MicroCal). Aliquots of ligands (2.5 μL) wereinjected into the 0.2-mL cell containing the protein solution in PBS buffer(pH 7.4) to achieve a complete binding isotherm. The resulting titrationcurves were fitted using MicroCal Origin software.

Evaluation of Kinetic Mechanism and Processing of Primary Data. Reactions ofA.17WT and its mutants (3–32 μM) or highly purified BChE (3–10 μM) fromhumanplasmawith1b (or 5) at concentrations ranging from10 to 500 μMwere

carried out in 0.1 M phosphate buffer (pH 7.4) at 25 °C. Stopped-flow meas-urementswithfluorescence detectionwere carried out using amodel SX.18MVstopped-flow spectrometer (Applied Photophysics). The Ab concentration wasfixed at 3 μM, and the concentration of 1b was varied from 1 to 50 μM. Nu-cleophile competition in the presence of NH2OH at concentrations rangingfrom 0 to 100 mM is carried out under the same reaction conditions withbackground correction.

ESI-FTICR-MS (Top Down). The A.17 antibody (10 μM) was incubated with 5(paraoxon; 2 mM in 0.1 M phosphate buffer, pH 7.4) at 37 °C for 16 h.Samples were analyzed directly using ESI-MS with a Bruker APEX Ultra FTICRmass spectrometer (Bruker Daltonics).

ACKNOWLEDGMENTS. We thank Professor Bernard Green for reading themanuscript and providing valuable comments and Professor Evgeny Nikolaev,Dr. Alexey Kononikhin, and Dr. Oleg Kharybin for performing ESI-FTICR-MSanalysis. Highly purified BChE from human plasma was kindly provided byProfessor Anikienko. This research was performed in frames of the Skolkovoprogram and supported by North Atlantic Treaty Organization SfP#982833,Russian Education Agency Contract N P1371, PICS Centre National de laRecherche Scientifique N 4238 France-Russia, Russian Foundation for BasicResearch Grants 10.04.00673a and 07.04.92168, Federal Special Purpose Pro-gram “Research and Development in Priority Directions of Growth of Scien-tific-Technological Complex of Russia for 2007–2012”, an International Centerfor Genetic Engineering and Biotechnology (Italy) grant, Scientific Schools64658.2010 “Chemical Basis of Biocatalysis,” and Programs of the RussianAcademy of Sciences “Fundamental Science for Medicine” and “Molecularand Cellular Biology.”

1. Benkovic SJ, Hammes-Schiffer S (2003) A perspective on enzyme catalysis. Science 301:1196e1202.

2. Tawfik DS (2010) Messy biology and the origins of evolutionary innovations. NatChem Biol 6:692e696.

3. Gupta RD, et al. (2011) Directed evolution of hydrolases for prevention of G-typenerve agent intoxication. Nat Chem Biol 7:120e125.

4. Belogurov A, Jr., Kozyr A, Ponomarenko N, Gabibov A (2009) Catalytic antibodies:Balancing between Dr. Jekyll and Mr. Hyde. Bioessays 31:1161e1171.

5. Lerner RA, Benkovic SJ, Schultz PG (1991) At the crossroads of chemistry and immu-nology: Catalytic antibodies. Science 252:659e667.

6. Savinov SN, Hirschmann R, Benkovic SJ, Smith AB, 3rd (2003) Investigation of an an-tibody-ligase. Evidence for strain-induced catalysis. Bioorg Med Chem Lett 13:1321e1324.

7. Stewart JD, Roberts VA, Thomas NR, Getzoff ED, Benkovic SJ (1994) Site-directedmutagenesis of a catalytic antibody: An arginine and a histidine residue play keyroles. Biochemistry 33:1994e2003.

8. Wagner J, Lerner RA, Barbas CF, 3rd (1995) Efficient aldolase catalytic antibodies thatuse the enamine mechanism of natural enzymes. Science 270:1797e1800.

9. Zhu X, et al. (2004) The origin of enantioselectivity in aldolase antibodies: Crystal struc-ture, site-directed mutagenesis, and computational analysis. J Mol Biol 343:1269e1280.

10. Barbas CF, 3rd, et al. (1997) Immune versus natural selection: Antibody aldolases withenzymic rates but broader scope. Science 278:2085e2092.

11. Demuth HU, et al. (1988) Molecular interactions in intermediate and transition statesin the self-stimulated inhibition of enzymes. Pharmazie 43:262e264.

12. Soumillion P, Fastrez J (2001) Novel concepts for selection of catalytic activity. CurrOpin Biotechnol 12:387e394.

13. Cesaro-Tadic S, et al. (2003) Turnover-based in vitro selection and evolution of bio-catalysts from a fully synthetic antibody library. Nat Biotechnol 21:679e685.

14. Reshetnyak AV, et al. (2007) Routes to covalent catalysis by reactive selection fornascent protein nucleophiles. J Am Chem Soc 129:16175e16182.

15. Kuzmic P (1996) Program DYNAFIT for the analysis of enzyme kinetic data: Applica-tion to HIV proteinase. Anal Biochem 237:260e273.

16. Oleksyszyn J, Boduszek B, Kam CM, Powers JC (1994) Novel amidine-containingpeptidyl phosphonates as irreversible inhibitors for blood coagulation and relatedserine proteases. J Med Chem 37:226e231.

17. Kim YR, et al. (2006) Heavy and light chain variable single domains of an anti-DNAbinding antibody hydrolyze both double- and single-stranded DNAs without se-quence specificity. J Biol Chem 281:15287e15295.

18. Schuermann JP, Prewitt SP, Davies C, Deutscher SL, Tanner JJ (2005) Evidence forstructural plasticity of heavy chain complementarity-determining region 3 in anti-body-ssDNA recognition. J Mol Biol 347:965e978.

19. Patten PA, et al. (1996) The immunological evolution of catalysis. Science 271:1086e1091.

20. Scotti C, Gherardi E (2006) Structural basis of affinity maturation of the TEPC15/Vkappa45.1 anti-2-phenyl-5-oxazolone antibodies. J Mol Biol 359:1161e1169.

21. Zhu X, Tanaka F, Lerner RA, Barbas CF, 3rd, Wilson IA (2009) Direct observation ofan enamine intermediate in amine catalysis. J Am Chem Soc 131:18206e18207.

22. Carletti E, et al. (2009) Structure-activity analysis of aging and reactivation of humanbutyrylcholinesterase inhibited by analogues of tabun. Biochem J 421:97e106.

23. Sanson B, et al. (2009) Crystallographic snapshots of nonaged and aged conjugates ofsoman with acetylcholinesterase, and of a ternary complex of the aged conjugatewith pralidoxime. J Med Chem 52:7593e7603.

24. Dvir H, Silman I, Harel M, Rosenberry TL, Sussman JL (2010) Acetylcholinesterase: From3D structure to function. Chem Biol Interact 187:10e22.

25. Foote J, Milstein C (1994) Conformational isomerism and the diversity of antibodies.Proc Natl Acad Sci USA 91:10370e10374.

26. Lindner AB, Eshhar Z, Tawfik DS (1999) Conformational changes affect binding andcatalysis by ester-hydrolysing antibodies. J Mol Biol 285:421e430.

27. James LC, Roversi P, Tawfik DS (2003) Antibody multispecificity mediated by confor-mational diversity. Science 299:1362e1367.

28. Wade H, Scanlan TS (2003) Binding and catalysis: A thermodynamic study on a cata-lytic antibody system. ChemBioChem 4:537e540.

29. Debord J, Harel M, Verneuil B, Bollinger JC, Dantoine T (2009) Microcalorimetric studyof the inhibition of butyrylcholinesterase by paraoxon. Anal Biochem 389:97e101.

30. Connelly PR, Thomson JA (1992) Heat capacity changes and hydrophobic interactionsin the binding of FK506 and rapamycin to the FK506 binding protein. Proc Natl AcadSci USA 89:4781e4785.

31. Jelesarov I, Bosshard HR (1999) Isothermal titration calorimetry and differentialscanning calorimetry as complementary tools to investigate the energetics of bio-molecular recognition. J Mol Recognit 12:3e18.

32. Hammes GG, Chang YC, Oas TG (2009) Conformational selection or induced fit: A fluxdescription of reaction mechanism. Proc Natl Acad Sci USA 106:13737e13741.

33. Becker EL (1967) The relationship of the structure of phosphonate esters to theirability to inhibit chymotrypsin, trypsin, acetylcholinesterase and C’Ia. Biochim BiophysActa 147:289e296.

34. Tramontano A, Ivanov B, Gololobov G, Paul S (2000) Inhibition and labeling ofenzymes and abzymes by phosphonate diesters. Appl Biochem Biotechnol 83:233e242.

35. Jackson DS, et al. (1998) Synthesis and evaluation of diphenyl phosphonate esters asinhibitors of the trypsin-like granzymes A and K and mast cell tryptase. J Med Chem41:2289e2301.

36. Boduszek B, et al. (1994) Dipeptide phosphonates as inhibitors of dipeptidyl pepti-dase IV. J Med Chem 37:3969e3976.

37. Griffiths AD, et al. (1994) Isolation of high affinity human antibodies directly fromlarge synthetic repertoires. EMBO J 13:3245e3260.

38. Yin J, Beuscher AE, 4th, Andryski SE, Stevens RC, Schultz PG (2003) Structural plasticityand the evolution of antibody affinity and specificity. J Mol Biol 330:651e656.

39. Wedemayer GJ, Patten PA, Wang LH, Schultz PG, Stevens RC (1997) Structural insightsinto the evolution of an antibody combining site. Science 276:1665e1669.

40. Henzler-Wildman KA, et al. (2007) Intrinsic motions along an enzymatic reactiontrajectory. Nature 450:838e844.

41. Jackson CJ, et al. (2009) Conformational sampling, catalysis, and evolution of thebacterial phosphotriesterase. Proc Natl Acad Sci USA 106:21631e21636.

Smirnov et al. PNAS | September 20, 2011 | vol. 108 | no. 38 | 15959

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

0