Immunological Origins of Binding and Catalysis in a Diels-Alderase Antibody

See allHide authors and affiliations

Science  20 Mar 1998:
Vol. 279, Issue 5358, pp. 1929-1933
DOI: 10.1126/science.279.5358.1929


The three-dimensional structure of an antibody (39-A11) that catalyzes a Diels-Alder reaction has been determined. The structure suggests that the antibody catalyzes this pericyclic reaction through a combination of packing and hydrogen-bonding interactions that control the relative geometries of the bound substrates and electronic distribution in the dienophile. A single somatic mutation, serine-91 of the light chain to valine, is largely responsible for the increase in affinity and catalytic activity of the affinity-matured antibody. Structural and functional studies of the germ-line precursor suggest that 39-A11 and related antibodies derive from a family of germ-line genes that have been selected throughout evolution for the ability of the encoded proteins to form a polyspecific combining site. Germ line–encoded antibodies of this type, which can rapidly evolve into high-affinity receptors for a broad range of structures, may help to expand the binding potential associated with the structural diversity of the primary antibody repertoire.

The immune system solves the problem of molecular recognition by generating a large library of structurally distinct antibodies and amplifying those with the requisite binding affinity and specificity in an affinity-based selection. By programming this system with chemical information about a reaction mechanism—for example, the structure of a putative transition state—one can examine the evolution of both binding energy and catalytic function (1). Functional and structural analysis of this process can provide insights into both the molecular basis for the remarkable efficacy of this combinatorial system and the mechanisms by which binding energy can be used to lower the activation energies of reactions (1-5). We now describe one such study of the antibody 39-A11 (6), which catalyzes a Diels-Alder reaction, a widely used and mechanistically well studied reaction in organic chemistry, but one that is rarely found in biological systems. The three-dimensional x-ray crystal structures of the 39-A11 Fab·hapten complex and of the germ-line precursor have been determined, and the immunological origins of this and related antibodies have been characterized.

Antibody 39-A11 was generated to the bicyclo[2.2.2]octene hapten 4, a mimic of the boatlike transition state of the Diels-Alder reaction. This antibody catalyzes the cycloaddition reaction of diene 1 and dienophile 2 to give the Diels-Alder adduct 3 (Scheme 1) (6). Structurally related haptens have been used to generate other antibodies that catalyze Diels-Alder reactions, suggesting that this is a relatively general design strategy (7, 8). Antibody 39-A11 was cloned and expressed as a humanized chimeric Fab (9), and the structure of the complex of the recombinant 39-A11 Fab fragment and hapten 4 was determined at 2.4 Å resolution (Fig. 1 and Table1).

Figure 1

(Left) Ribbon superposition of the variable regions of the germ-line Fab without hapten (purple) and the mature Fab·hapten 4 complex (red). The side chains of the somatic mutation sites are indicated for the germ-line and mature antibodies: ValL27c→Leu and SerL91→Val. (Right) Close-up view of the Diels-Alder 39-A11 active site with the transition state analog bound to the mature form of the antibody. The mature antibody is shown in red, and the germ-line antibody without hapten in purple. No significant differences are apparent between the two structures. The hapten molecule is shown with carbon atoms in green, oxygen in red, and nitrogen atoms in blue. (Bottom) Stereoview of the 1F o − 1F c electron density surrounding the transition state analog (F o and F care the observed and calculated structure factors, respectively).

Figure 2

Sequences of the VH and VLregions of the structurally related antibodies 39-A11, DB3, and TE33, and of the germ-line precursors VMS9 and Vκ1A, respectively. Vκ1A and VMS9 were identified as likely VL and VH germ-line candidates, respectively, through a homology search of the Kabat database (20). On the basis of their published sequences, 5′ PCR primers were designed to anneal to upstream untranslated DNA of the Vκ1A and VMS9 genes. PCR amplification with 39-A11 hybridoma DNA as template in conjunction with 3′ light and heavy J region–specific primers yielded several clones of both chains, which were then sequenced. In each instance, the flanking regions were shown to be identical to the published Vκ1A and VMS9 sequences over a region of hybridoma DNA sufficiently large to identify the rearranged gene (∼500 and 400 nucleotides for the light and heavy chain genes, respectively). Abbreviations for the amino acid residues are: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. Dashes represent residues identical to the corresponding Vκ1A or VMS9 sequence.

Figure 3

Structures of ligands used in binding assays with 39-A11 and its germ-line precursor.

Figure 4

Superposition of the CDRL3 and CDRH3 loops of antibodies DB3, TE33, and 39-A11 with bound steroid (green), peptide (blue), and hapten 4 (purple), respectively. TrpH50, Asn/SerH35, and Trp/ArgH100are also shown.

Scheme 1
Table 1

Data collection and refinement statistics for the mature (with hapten) and germ-line (without hapten) Fabs (31).

View this table:

Well-defined density for the hapten was observed in the 1 Fo − 1Fc omit map (Fig.1). The hapten is bound in a cleft ∼9 Å wide and ∼12 Å deep, with ∼194 Å2 of the hapten surface (79% of the total solvent-accessible surface excluding the linker arm) buried within the Fab. There are 89 van der Waals interactions and two hydrogen bonds between the hapten and antibody, with most of these contacting the heavy chain. The bicyclo[2.2.2]octene moiety of hapten 4, which corresponds to the cyclic 4+2 π electron system of the transition state, is buried in a hydrophobic pocket, free of solvation. The walls of this cavity consist of the side chains of residues PheH100b [antibody nomenclature described in (20)], AsnH35, TrpH47, ValL91, ProL96, GlyH33, TrpH50, AlaH95, and ArgH100 (where H and L represent heavy and light chains of the antibody, respectively). The carbonyl oxygen of the carbamate moiety at the bridgehead position of 4 (the C1 substituent in the diene) is hydrogen-bonded through a water molecule to Nε1 of TrpH50. The succinimido moiety of 4, which corresponds to the maleimide group of the dienophile, is involved in a π-stacking interaction with TrpH50. The N -phenyl substituent is packed against the backbone of residues GlyH33 and GluH96 and the methylene groups of ArgH97. The succinimido carbonyl group anti to the carbamate moiety of 4 is hydrogen-bonded to the side chain amide of AsnH35, which is oriented by a hydrogen bond between the carboxamide oxygen and the side chain of TrpH47.

The x-ray crystal structure of the Fab·hapten 4complex suggests that antibody 39-A11 binds the diene and dienophile in a reactive orientation and reduces translational and rotational degrees of freedom (10, 11). The dienophile is oriented by hydrogen-bonding and π-stacking interactions with the maleimide ring. The diene is bound in a hydrophobic pocket in close proximity to the dienophile, with the position of the carbamate substituent fixed by a water-mediated hydrogen bond to TrpH50 (Fig. 1). Although the carbon-carbon bond lengths of bicyclo[2.2.2]octene are shorter than those that would be present in either a synchronous or nonsynchronous transition state, it appears that both transition states can be accommodated in the active site. In addition to promixity effects, the energetics of the antibody-catalyzed reaction may be influenced by a hydrogen bond between the side chain carboxamide group of AsnH35 and the maleimide group. This interaction should render the olefin more electron deficient and, as a result, a more reactive dienophile (Fig. 1) (12). In the instance of an asymmetric dienophile, containing only one electron-withdrawing carbonyl group, this interaction might be expected to enhance the formation of the disfavored regioisomer of the Diels-Alder adduct.

The structure of the 39-A11·hapten 4 complex provides an explanation for the ∼1000 times greater binding affinity of 39-A11 for hapten 4 than for product 3. Conformational constraints imposed by the [2.2.2]-bicyclic framework lock the cyclohexene ring of the hapten into a boatlike geometry distinct from that of the product. These constraints result in less favorable van der Waals and hydrogen-bonding interactions between active site residues and substituents on product 3 than between active site residues and hapten 4, for example, the water-mediated hydrogen bond between TrpH50 and the carbamate moiety of the linker. The crystal structure also suggests that mutation of the active site residues ProL96 and ValL91 in 39-A11 to residues with increased hydrophobic surface area might lead to an increase in rate as a result of improved packing interactions with the kinetically favored endo transition state (the structure of hapten 4 accommodates both endo and exo transition states). Indeed, three out of six site-directed mutations of these residues (ValL91 to Tyr, and ProL96 to Phe or Tyr) were associated with about 5- to 10-fold increases in the catalytic rate constant ( kcat).

The structure of antibody 39-A11 can be compared with those of other proteins that catalyze pericyclic reactions. In the case of the antibody 1F7 (5) and the Bacillus subtilis(13), Escherichia coli (14), and yeast (15) chorismate mutases, which catalyze the Claisen rearrangement of chorismate to prephenate, the protein also appears to organize the substrate into a reactive conformation by a network of hydrogen-bonding and van der Waals interactions (10, 11). In the enzyme active sites, there are also hydrogen bonds to the enol ether oxygen of chorismate, which have been postulated to stabilize the developing charge at this center in the transition state (16). Nuclear magnetic resonance and x-ray crystallographic analyses of monoclonal antibody AZ-28, which catalyzes a related pericyclic rearrangement (oxy-Cope), indicate that this antibody also stabilizes the conformationally restricted cyclic transition state (3). In addition, AZ-28 may increase the reaction rate by enhancing the extent of electron density on the hydroxyl substituent of the substrate through hydrogen-bonding interactions. Thus, it appears that catalysis of these pericyclic transformations involves both restriction of rotational and translation entropy in the substrate as well as hydrogen-bonding interactions that modulate electron densities on key substituents in the transition state.

Analysis of the germ-line precursor to antibody 39-A11 provides an opportunity to examine the evolution of this biological catalyst in a combinatorial system that reflects features of the natural evolutionary process. The VL gene for 39-A11 exhibits three nucleotide differences (one of which is silent) relative to its germ-line precursor Vκ1A (Fig.2) (17); the V region is joined in frame to Jκ1, resulting in a proline at junctional position 96. The VH gene differs by one nucleotide (which is silent) from its germ-line precursor VMS9, a member of the VGAM3.8 VH family (18). There may also be one difference in the last codon of the gene, but this nucleotide is likely not encoded by the variable region and therefore probably does not represent a somatic mutation; D region DSP2.2 (19) and JH4 (20) genes were used in unmutated form (Fig. 2). This analysis indicates that affinity maturation of antibody 39-A11 results in only two somatic mutations: a Val→Leu substitution at position 27c in CDRL1, and a Ser→Val substitution at position 91 in CDRL3. The functional consequences of affinity maturation on binding affinity and catalysis were determined by expressing and characterizing the germ-line antibody as well as the individual somatic mutants. The dissociation constant ( Kd) of the germ-line antibody for hapten4 is 379 ± 16 nM (21). The two somatic mutations result in a 40-fold increase in binding affinity ( Kd = 10 ± 0.3 nM) of the mature antibody 39-A11 for hapten 4, with virtually all of this increase associated with the somatic mutation at position L91. The effects of the somatic mutations on kcat and the Michaelis constant ( Km) parallel their effects on binding affinity. The kcat, Km(1), and Km(2) values for the recombinant Fab 39-A11 are 0.67 s−1, 1200 μM, and 740 μM, respectively, whereas the corresponding values for the germ-line antibody are 0.17 s−1, 1400 μM, and 450 μM (21). The effects of affinity maturation on catalysis, which are reflected primarily in the values of kcatand Km(1), again result largely from the SerL91→Val somatic mutation.

To probe the structural consequences of affinity maturation, we determined the three-dimensional structure of the germ-line antibody Fab at 2.1 Å resolution (Fig. 1 and Table 1). The structures of the unliganded germ-line antibody and the 39-A11·hapten complex are very similar. The root-mean-square deviations for the Cα positions of VH and VL are 0.45 and 0.51, respectively; the overall deviation for the variable domain Cα positions is 0.71. The only relatively large structural change is in the position of the poorly defined side chain of PheL87, which is distant from the active site. Smaller differences are apparent at the site of the SerL91→Val somatic mutation (the distance from ValL91 to the unsubstituted bridgehead carbon of4 is 4 Å) and the nearby residue HisL34. Nonetheless, a comparison of the two structures indicates that neither somatic mutation nor ligand binding results in substantial structural or conformational changes in the active site.

Thus, the affinity maturation of antibody 39-A11 presents an opportunity to examine a solution to the problem of molecular recognition markedly different from that involving the esterolytic antibody 48G7, which we previously described (2). In the latter instance, nine somatic mutations, none directly contacting the hapten, contribute additively to the 30,000-fold increase in affinity for the nitrophenyl phosphonate transition state analog ( Kd = 4 nM). Structural changes occur on binding of hapten to the germ-line antibody that result in enhanced antibody-hapten complementarity. These structural changes were further optimized by affinity maturation, resulting in “lock and key binding” of hapten to the mature antibody. In contrast, the germ-line precursor to antibody 39-A11 appears to be a better start point—only one somatic mutation in the combining site is required to bind hapten with an affinity similar to that of 48G7. This difference may be a consequence of the hydrophobic nature of hapten 4; alternatively, antibody 39-A11 may have evolved from a relatively polyspecific combining site that was selected in the germ-line repertoire for its ability to bind a structurally diverse array of antigens. Therefore, we assayed binding of both 39-A11 and its germline precursor to a panel of chemically defined, structurally diverse haptens conjugated to bovine serum albumin (BSA) (Table2 and Fig.3).

Table 2

Dissociation constants for the binding of ligands to antibody 39-A11 and its germ-line precursor.

View this table:

Both antibodies bind nine haptens containing a broad range of hydrogen-bonding, charged, and hydrophobic groups [in contrast, the nitrophenyl phosphonate–specific antibody 48G7 showed no cross-reactivity when screened against the same panel of hapten-BSA conjugates (2)]. Comparison of these affinities with those for hapten 4 provides a measure of the polyspecificity. The germ-line precursor of 39-A11 shows affinities for haptens Athrough I that are roughly within an order of magnitude to that for its own hapten 4, whereas the affinity of 39-A11 for hapten 4 is up to 1000 times those for haptensA through I. These results suggest that the germ line–encoded antibody is polyspecific and can be selected for clonal expansion and subsequent affinity maturation by a wide variety of antigens, including those shown in Fig. 3. This polyspecificity may be general to several germ line–encoded antibodies and may have been selected for by the immune system to provide a mechanism for rapid generation of antibodies of moderate to high affinity for a broad range of antigens.

Three other antibodies (22-24) that were clonally selected on the basis of their intrinsic affinities for markedly different ligands use VH and VL chains highly homologous to those of 39-A11. Antibodies DB3 (22), TE33 (23), and IE9 (24) were raised against progesterone, a 16–amino acid peptide, and a hexachloronorbornene derivative, respectively. The three-dimensional crystal structures of DB3 (22) and TE33 (23) have been solved and, together with the structure of 39-A11, make possible a detailed structural analysis of how an antibody of limited diversity is able to bind a variety of structurally distinct antigens (Fig.4). Antibody DB3 is specific for progesterone and most likely comprises Vκ1A, Jκ1, VMS9, and JH4 (25); it also shows cross-reactivity with various structurally related progesterone analogs (26). Both DB3 and IE9 show some cross-reactivity (24). Antibody TE33 is specific for the cholera toxin peptide VEVPGSQHIDSQKKA, and most likely comprises Vκ1C, Jκ4, V264 (a member of the VGAM3.8 family), and JH1 (27). Most of the differences in sequence among these antibodies are located in CDRH3 and are not germ line–encoded (Fig. 2). All three antibodies use a light chain variable region encoded by the Vκ1 gene, which is common to a relatively large population of antibodies that bind a large number of antigens including proteins, DNA, steroids, peptides, and small haptens (17).

The combining sites of 39-A11, DB3, and TE33 present a large, highly conserved binding surface formed predominantly by the CDRL1, CDRL3, and CDRH3 hypervariable loops; the CDRH3 and CDRL3 loops together with TrpH50 form a deep hydrophobic binding pocket (Fig. 4). A more shallow region of the binding pocket is dominated by VL contacts, with CDRL3 providing the floor and CDRL1 bordering the pocket. In antibodies DB3 and 39-A11, TrpH50and residue H100 in the CDRH3 loop sandwich the hapten, providing critical hydrogen-bonding or hydrophobic contacts that define opposite walls of the deep binding pocket—TrpH100 in DB3 packs with the central nonpolar region of the steroid, and ArgH100 and TrpH50 in 39-A11 provide key hydrophobic and hydrogen-bonding interactions with hapten 4. In these same two antibodies, the amide side chain of AsnH35 at the bottom of the binding pocket is positioned to donate a key hydrogen bond to the ligand. The conserved length of CDRH3 and the conserved residue in position H100 result from antigen-driven selection during recombination. The length of the CDRH3 loop of IE9 is also conserved, and this antibody contains arginine at position H100. A similar combining site is used by TE33 to bind its peptide antigen. Four NH2-terminal residues of the peptide form a critical portion of an antiparallel β turn that is buried in the binding pocket. These residues are oriented in the same manner as are hapten4 and progesterone in their respective binding pockets (Fig.4). Again, the indole side chain of TrpH50 plays a prominent role by packing against Gly5 of the peptide. However, TE33 has a shorter CDRH3 loop (only seven amino acids), allowing the COOH-terminus of the peptide to exit the binding pocket.

Structural analysis of the esterolytic antibody 48G7 and its germ-line precursor suggested that, in addition to sequence diversity, conformational diversity intrinsic to some germline antibodies may contribute to the ability of the germ-line repertoire to bind such a wide array of chemical structures (2). The above structural analysis suggests that another important mechanism may involve the selection through evolution of a set of germ-line antibodies that are polyspecific—a feature shared by the major histocompatibility complex molecules, which bind many diverse peptides in the cellular immune response (28). Germ-line gene duplication may have created sets of closely related genes (Vκ1 and VGAM3.8) whose products contain subtle, but mechanistically important, amino acid differences that, in conjunction with CDRH3 loop diversity, may allow presentation of several variations of the combining site. This combining site may have been selected during evolution as an optimal start point for rapid evolution of high-affinity, specific combining sites for a broad range of structures through a limited number of somatic mutations (these mutations may also remove interactions with idiotypic antibodies that regulate self recognition). Thus, the immune system likely relies on a variety of strategies, including conformational diversity and polyspecificity, in addition to somatic processes, to solve the problem of molecular recognition.

  • * These authors contributed equally to this work.


View Abstract

Navigate This Article