A structural basis for the activity of retro-Diels–Alder catalytic antibodies: Evidence for a catalytic aromatic residue

Abstract

The nitroxyl synthase catalytic antibodies 10F11, 9D9, and 27C5 catalyze the release of nitroxyl from a bicyclic pro-drug by accelerating a retro-Diels–Alder reaction. The Fabs (antigen-binding fragments) of these three catalytic antibodies were cloned and sequenced. Fab 9D9 was crystallized in the apo-form and in complex with one transition state analogue of the reaction. Crystal structures of Fab 10F11 in complex with ligands mimicking substrate, transition state, and product have been determined at resolutions ranging from 1.8 to 2.3 Å. Antibodies 9D9 and 10F11 show increased shape complementarity (as quantified by the program sc) to the hapten and to a modeled transition state as compared with substrate and product. The shape complementarity is mediated to a large extent by an aromatic residue (tyrosine or tryptophan) at the bottom of the hydrophobic active pocket, which undergoes π-stacking interactions with the aromatic rings of the ligands. Another factor contributing to the different reactivity of the regioisomers probably arises because of hydrogen-bonding interactions between the nitroxyl bridge and the backbone amide of PheH101 and possibly a conserved water molecule.

Antibodies have been shown to catalyze a wide variety of reactions with exquisite control over the reaction pathway (1–3). The guiding paradigm for designing catalytic antibodies has been to induce a binding pocket complementary to a given reaction's transition state (4). Thus, immunization is carried out against a hapten with a molecular shape and charge distribution resembling those of the transition state as closely as possible. Although charges play a key role to mimic transition states of ionic reactions such as acidic or basic hydrolytic reactions, the shape of the transition state analogue has been particularly critical for the Diels–Alder reaction, for which several catalytic antibodies have been reported (5–10) and crystallized (11–15). Aromatic residues have been observed recurrently in the binding pockets of Diels–Alderase antibodies. We have used x-ray crystallography to identify a similar aromatic residue in the binding pocket of the retro-Diels–Alder catalytic antibodies 10F11 (16) and 9D9 (17). This aromatic residue is conserved as well in the third retro-Diels–Alder antibody 27C5. The analysis based on crystal structures of complexes with substrate, transition state analog, and product suggests that this aromatic residue mediates selective shape complementarity to the reaction's transition state and its analogues over substrate and product and contributes in this way to catalysis.

Materials and Methods

Fab Preparation and Purification.

Monoclonal catalytic antibodies 10F11, 9D9, and 27C5 were produced from their respective hybridoma cell lines by cell culture and purified to homogeneity by ammonium sulfate precipitation, ion exchange, and protein G chromatography. The Fabs 10F11 and 9D9 were generated by papain digestion as described by Porter (18) using immobilized papain (Sigma). Fab fragments were separated from undigested antibodies and Fc fragments by ion exchange chromatography on a Mono Q FPLC column (Resource 6 ml, Amersham Pharmacia) by using 50 mM Tris, pH 7.8/0–350 mM NaCl.

Ligand Preparation.

Hapten 3 and 4 and reaction product 2 were prepared as described before (5, 6). Ligand 6 was obtained by reaction of 2b with singlet oxygen (19) as follows: 9-[2′-carboxylethyl]-10-methylanthracene endoperoxide (6). Dropwise aqueous H₂O₂ (5 equivalent) was added to a solution of 2b (20 mg, 0.076 mmol) in 2 ml of methanol and aqueous NaClO (0.32 ml, 10 equivalent). The reaction was stirred for 4 h at 25°C, and the product was purified directly by preparative reverse-phase HPLC (RP C-18, acetonitrile-water). Lyophilization of the main fraction gave 2b (4.7 mg, 21%) as a pale yellow solid, ¹H NMR (400 MHz, CDCl₃): δ (ppm): 7.39 (m, 4H), 7.29 (m, 4H), 3.10 (t, 2H, J = 7.6Hz), 3.08 (t, 2H, J = 7.6Hz), 2.90 (s, 3H); ¹³C NMR (75 MHz, CDCl₃): δ (ppm) 177.8, 141,3, 140.0, 127.4, 127.3, 121.1, 121.0, 80.5, 79.4, 27.8, 23.3, 13.6; MS: 297 (MH⁺).

Transition State Modeling.

Transition state 7 was calculated starting from the energy minimized structure of substrate 2c by enforcing the transition state bond length reported by Houk (20) for the reaction of nitroxyl with butadiene. The transition state was located by using semiempirical calculations with SPARTAN 5.1.

Cloning and Sequencing of the Recombinant Fab Fragments 9D9, 10F11, and 27C5.

The extraction of total RNA was done following the protocol of TRI Reagent-RNA isolation reagent (catalog no. TR118, Molecular Research Center, Cincinnati). mRNA was extracted from 2 × 10⁸ hybridoma cells by using a phenol-chloroform guanidium isothiocyanate procedure. About 5–10 μg total mRNA was annealed to 1,000 ng of an oligo(dT) primer. After reverse transcription using 400 units of Moloney Murine Leukemia Virus Reverse Transcriptase (Boehringer Mannheim), heavy and light chains were amplified by using DNA polymerase Vent (New England Biolabs). The primer sequences used are: (i) 5′-GCGACTGAGCTCGATGTTTTGTTGACCCAGAC-3′; (ii) 5′-CAGTGCTGTCGACGAGCTGAAGCTGGTGGAGAC-3′; (iii) 5′-GCGCCGTCTAGAACACTCATTCCTGTTGAAGC-3′; and (iv) 5′-GTTCTGACTAGTCTAGGGCACTCTGGGCTCAAT-3′. After purification (Qiagen Gel Extraction kit), PCR products were cut with XbaI and SacI for V_L–C_L and with SalI and SpeI for V_H–C_H1 and ligated into the pComb3 vector (21). After transformation, positive clones were confirmed by double-strand sequencing (Microsynth, Balgach, Switzerland).

Crystallization.

Crystallizations were performed by using the vapor diffusion method. The Fab fragment 10F11 (16–20 mg/ml) with hapten (0.6 mM) in 10 mM Tris⋅HCl, pH 7.8, was mixed with 1 μl of buffer (4: 0.05 M imidazole, pH 7.0; 6: 0.05 M glycine, pH 9.0; 2b: 0.05 M citric acid, pH 3.0; 3: Tris⋅HCl, pH 6.5) and 2 μl of precipitant [12–18% (wt/vol) polyethylene glycol 3550]. Fab 9D9 drops consisting of a 3-μl aliquot of the protein solution (15 mg Fab per ml in 10 mM Tris⋅HCl, pH 7.8) were mixed with 1 μl of buffer [0.2 M ammonium sulfate/0.1 M sodium acetate, pH 4.6/30% (wt/vol) PEG 4000]. All crystallizations were carried out at 20°C.

X-Ray Data Collection and Structure Determination.

Data collection crystals were flash-cooled to 110 K by using 28% PEG 400 as cryoprotectant. Data were processed by using the programs MOSFLM 6.0 and scala of the CCP4 program suite (22). Data collection statistics are given in Table 1. Crystals of 10F11 were pseudomerohedrally twinned as indicated by the cumulative intensity statistics output by truncate (22). Before structure solution by molecular replacement (effected with the program cns; ref. 23), data were detwinned by using a local version of detwin from the CCP4 package (22). The Fab fragment 28B4 (PDB entry code 1KEL) proved to be a suitable search model (24). The atomic model was rebuilt by using the program o (25). Refinement was carried out with cns version 1.0 employing the option to input twinned data together with an estimate of the twinning parameter. Fab residue numbering follows the Kabat numbering scheme throughout (26).

Table 1.

Data collection and refinement statistics

	Fab 10F11⋅6	Fab 10F11⋅4	Fab 10F11⋅3	Fab 10F11⋅2b	Fab 9D9 apo
Data collection
Space group	P21	P21	P21	P21	P212121
Unit cell	a = 40.52 Å, b = 140.09 Å, c = 85.478 Å, β ∼ 90°	a = 40.63 Å, b = 139.89 Å, c = 85.20 Å, β ∼ 90°	a = 40.66 Å, b = 140.15 Å, c = 85.33 Å, β ∼ 90°	a = 40.384 Å, b = 139.768 Å, c = 84.86 Å°, β ∼ 90°	a = 48.71 Å, b = 80.36 Å, c = 125.10 Å
Resolution range, Å	30.8–2.0	30.3–1.77	30.7–2.0	30.5–2.3	40.0–2.4
Outer resolution shell	2.11–2.0	1.85–1.77	2.03–2.0	2.42–2.3	2.53–2.4
Observations	721,745	1,025,225	649,910	393,333	182,312
Unique reflections	64,582	74,192	78,457	39,582	19,718
Completeness (%)	99.5 (97.5)	81.7 (95.9)	99.8 (97.5)	89.5 (83.2)	99.0 (98.3)
Mean I/σ	26.87 (7.3)	24.35 (9.9)	16.61 (8.3)	16.7 (3.0)	30.2 (10.6)
Rsym*	0.046 (0.138)	0.051 (0.136)	0.037 (0.097)	0.063 (0.337)	0.027 (0.116)
Refinement statistics
Twinning fraction	0.37	0.38	0.33	0.38	–
Atoms/water molecules	6780/93	6786/58	6784/62	6776/44	3368/19
Rcryst, %†	20.83	20.4	20.8	18.8	25.0
Rfree, %†	24.8	25.6	26.1	25.1	31.5

^*R_sym = ∑_hkl∑_i|I(hkl; i) − 〈I(hkl)〉|/∑_hkl∑_i〈I(hkl)〉. 

^† R/R_free = ∑_hkl|F_obs(hkl) − F_calc(hkl)|/∑_hklF_obs(hkl), where summations are carried out separately over working and test sets, respectively. 

Results and Discussion

Antigen-Binding Pocket.

The structure of Fab 10F11 was solved as complexes with four different ligands: substrate analog 6, haptens 3 and 4, and product analog 2b (Fig. 1). The structure of Fab 9D9 was solved in the apo-form and as a complex with hapten 3, although the latter crystals were of inferior quality and will not be discussed further here. In case of 10F11, the bound ligands are clearly visible in the electron density maps, allowing unequivocal identification of the catalytic pocket (Fig. 2), which is mainly hydrophobic as in other Diels–Alderase antibodies (11–15). The ligands are deeply buried in this pocket, which is formed by Tyr^H53, Tyr^H58, Ser^H100, Phe^H101, Trp^H104/Tyr^H104, Tyr^L37, Gly^L96, and Phe^L99 (Fig. 2). These residues are conserved in the three catalytic retro-Diels–Alderase antibodies, 9D9, 10F11, and 27C5 (Fig. 3). Contrary to Diels–Alderase antibody 1E9 (11), where the fit between protein and ligand is so snug that no interfacial cavities are discernable, the binding pockets of 9D9 and 10F11 seem large enough (11 Å deep, 14 Å wide) to accommodate two to three water molecules beside the ligands. Ligand recognition is achieved through multiple van der Waals contacts. A rather striking π-stacking interaction can be observed between the aromatic rings from all ligands and the aromaticside chain of Trp^H104, in the case of Fab 10F11, or Tyr^H104, in the case of Fab 9D9. Residue H104 is held firmly in place by a hydrogen-bonding network; the completely buried side chain of Glu^L39 accepts two hydrogen bonds from the main chain amides of residues H104 and H105. The NH group of the indole ring (10F11) or the phenolic OH group (9D9) makes a hydrogen bond to the carbonyl oxygen of Gly^L96. The polar side chain of the ligands point out of the pocket toward the solvent.

Figure 1.

Reaction scheme and haptens used in this study. The retro-Diels–Alder reaction of pro-drug 1a (k_un = 9.8 10⁻⁶ s⁻¹ conditions: 31°C in PBS pH 7.4, 10% vol/vol dimethylformamide) releases nitroxyl and anthracene 2a. The reaction is catalyzed by antibodies 9D9 (anti-3), 10F11 (anti-5), and 27C5 (anti-3,4). The reaction with the regioisomeric substrate 1b (k_un = 7.1 10⁻⁶ s⁻¹) is also catalyzed weakly by the antibodies. The unreactive substrate analog 6 and product analogue 2b were used for crystallization. Substrate analogue 1c and product analog 2c and transition state model 7 were used for docking studies.

Figure 2.

Sequence alignment and comparison of the CDR H (A) and CDR L (B) corresponding to the three nitroxyl-synthase catalytic antibodies 9D9, 10F11, and 27C5, the Diels–Alderase antibodies 1E9 and 39-A11, and the noncatalytic antibody Mrk-16. Identical residues are shown in red, similar residues are shown in blue, and different residues are shown in black. Black stars represent residues in close contact with the hapten.

Figure 3.

Active site of the abzyme. (Left) 10F11 in complex with OXY (compound 6). The hydrogen bonding network is shown as dotted lines. (Right) Surface representation of the binding pocket of Fab 10F11 with the transition state analogue BCN (compound 4) of the retro-Diels–Alder reaction. View is from the top of the binding site showing the fit of the ligand BCN. The light pink transparent surface represents V_L and the light blue transparent surface represents V_H. Note the presence of water molecules inside the pocket. Figure prepared with dino (http://www.dino3d.org).

Database searches revealed another antibody with a very high sequence identity (86%) to the retro-Diels–Alderases described here. This antibody, Mrk-16 (PDB ID code 1BLN), was originally raised against P glycoprotein (27). Strikingly, at position H104, Mrk-16 also has a tryptophan residue. However, to be catalytically active in the retro-Diels–Alder reaction, at least two mutations should be necessary: Y^H98 → G and T^H33 → G (numbering as in 10F11) to open the pocket and make space for the substrate to stack with the putative catalytic tryptophan, which in the Mrk-16 structure interacts in a Τ stack manner with Y^H98.

Shape Complementarity.

Given the fact that the noncatalyzed retro-Diels–Alder reaction of 1a/b is insensitive to medium effects (the rate increases by only threefold when going from water to 95% dimethylformamide, and is similarly slow for compound 1c in organic solvents) (28, 29), antibody catalysis must be interpreted in terms of specific interactions within the binding pocket. The structural analogies between haptens 3, 4, and 5 and the reaction's transition state 7 are quite strong, particularly with regard to the dihedral angle θ between the two aromatic rings of the anthracene or acridine moiety. This angle is close to 150° in these molecules, in between that of the substrate 1 (θ = 118°) and the product 2 (θ = 180°). This finding suggests that shape complementarity of the catalytic pocket to the transition state of the reaction might be operating.

The program sc (30) was used to compute the shape complementarity statistics index (Sc). This statistics index measures the geometric surface complementarity via the use of normal products, and the extent to which the interacting surface elements are brought into proximity via an exponential distance separation term. Sc values usually range from 0.70 to 0.76 for proteinase–protein inhibitor surfaces, from 0.64 to 0.68 for antigen–antibody complexes, and from 0.45 to 0.70 for MHC–peptide complexes (30).

For 10F11, Sc was computed for the experimentally determined complexes of substrate analogue 6, transition state analog 4, and reaction product 2b, as well as for docked complexes of model substrate 1c, its calculated transition state 7, and the reaction product 2c (Fig. 1). Care was taken to dock the ligands into those Fab conformations derived from the experimental structures of the most closely related complexes to take into account as much as possible the induced fit of the CDR-H3 (see below). In the case of Fab 9D9, docking had to be performed into the apo-structure because no reasonably well diffracting crystals of complexes could be obtained, and hence no conformational rearrangements of the Fab could be taken into account. In both cases, the analysis revealed that the tightest fit to the antibody is achieved by the transition state analogue 4 and the calculated transition state 7 over all other ligands (Table 2). Very similar results were obtained by using the different shape-complementarity analysis program suite FADE/PADRE (31, 32).

Table 2.

Shape complementarity statistics

Compound	10F11	9D9*	10F11–WH104Y†
6	0.822  ± 0.02‡	0.699	0.713
4	0.857  ± 0.003‡	0.722	0.725
3	0.762  ± 0.004‡	0.659	–
2b	0.786  ± 0.01‡	0.614	–
1c fast§	0.796	0.659	0.672
1c slow§	0.755	0.657	0.645
7 fast¶	0.848	0.667	0.704
7 slow¶	0.841	0.675	0.691
2c	0.748	0.570	–

^*All ligands were docked into the 9D9 apo-structure with no remodeling of the loop H100–H104. 

^† Modeled 10F11 mutant Trp^H104 → Tyr. 

^‡ Experimental structures. Sc value averaged over the two crystallographically independent molecules ± SD. 

^§ Modeled substrate 1c docked in fast and slow orientation. 

^¶ Modeled transition state 7 docked in the fast and slow reacting orientation. 

A small conformational change of the loop H99-H102 can be observed in the 10F11 structures going from the substrate analogue 6 toward the product 2b (Fig. 4 Left). This segment of CDR-H3 moves toward the hapten because of the space becoming available when the O—O (or HN—O in case of the true substrate) bridge gets removed. The Sc analysis revealed that this induced fit movement of loop H99–H102 contributes to shape complementarity, and thus possibly to catalysis. For example, docking the transition state 7 into the Fab structure derived from complex 10F11⋅6 resulted in shape complementarity scores of 0.799 and 0.804 for the faster and slower reacting substrate, respectively, whereas docking the model product 2c resulted in a score of 0.665, all these values being significantly lower than for the true experimental structures.

Figure 4.

(Left) Overlay of 10F11⋅6 (green)“OXY,” 10F11⋅4 (yellow) “BCN,” and 10F11⋅2b (magenta) “ANT.” The small movement of the flap H99–H102 is clearly visible at Ser-H100. Water molecules S1 and S26 are shown as small spheres, and water molecule S127, which is only present in the 10.F11⋅6 structure, is shown as large cyan sphere. (Right) Overlay of 10F11 with 9D9. 9D9 is shown in orange, 10F11 is shown in cyan. Amino acids labeled in black represent the conserved residues, and red labels indicate the exchanges of amino acids at equivalent positions.

Role of Residue H104 in Catalysis.

The increased shape complementarity of antibody 10F11 to the transition state 7 and hapten 4 furthermore correlates with the higher catalytic efficiency of 10F11 over 9D9 if one takes not only the higher absolute number but also the differential between substrate and transition state into account (Fig. 4 Right). The higher Sc scores of 10F11 originate in the larger side chain of Trp^H104 as compared with Tyr^H104. This finding was demonstrated by remodeling this residue in the 10F11⋅4 structure as a tyrosine, adjusting the conformation to the one observed in 9D9. The Sc score obtained for this theoretical mutant are similar to those derived from Fab 9D9, with the transition state 7 scoring highest in the orientation of the fast reacting substrate.

Despite of the strong correlation observed between shape complementarity and catalysis one could also envision that other interactions mediate catalysis directly. The so-called normal Diels–Alder reactions proceed fastest between electron-rich dienes [energetically high HOMO (highest occupied molecular orbital)] and electron-deficient dienophiles [energetically low LUMO (lowest unoccupied molecular orbital )]. The π-stacking interaction between the electron-rich aromatic side chain of residue H104 and the diene moiety of the substrate could stabilize the reaction's transition state by raising the energy of the diene's HOMO. However, similar π-stacking interactions between a tryptophan side chain (Trp^H50) and the corresponding haptens have been reported in Diels–Alderase antibodies 1E9 and 39-A11, and in those cases the indole ring stacks onto the portion of the hapten corresponding to the dienophile of the Diels–Alder reaction. Here, an electron-donating interaction between a tryptophan and a dienophile would increase the HOMO–LUMO energy gap and thus slow down the reaction. The positioning of the tryptophan side chain relative to the reaction's transition states of Diels–Alderase antibodies, thus, does not necessarily depend on possible orbital interactions during the reaction, and its role as mediator of shape complementarity as shown above seems to be more likely.

It must be stressed again that the modeled structures are inherently less accurate because of the neglect of conformational changes of the antibodies and because of the less accurately known ligand positions. An explanation of the different reactivities of the two regioisomers 1a/1b using the shape complementarity argument alone fails, as can be seen from Table 2. The differential in the Sc values between educt and transition state is higher for the slowly reacting regioisomer; this could be a modeling artifact arising from the factors explained above. Alternatively, other interactions should be taken into account as well to explain the different reactivities of compounds 1a and 1b, as discussed below.

Hydrogen Bonding Interactions.

The normal Diels–Alder reactions can be accelerated by Lewis acids, which complex to the dienophile and hence lower its LUMO energy further. In the10F11⋅6 complex a hydrogen bond is observed between the bridge oxygen O² of 6 and the amide NH of Phe^H101 of the Fab. Similarly, water molecule S127, which is found in the active site, is placed at a key position to create three hydrogen bonds, namely with the O² of 6 (3.0 Å), and with the main chain amide nitrogen atoms of Ser^H100 (3.1 Å) and Tyr^H53 (3.1 Å).

These hydrogen bonds might contribute to catalysis by lowering the LUMO energy of the dienophile when the favored substrate 1a enters a catalytically productive hydrogen-bonding interaction on the bridge's oxygen atom, as is found in the case of 6. A similar catalytic role of hydrogen bonds has been argued for Diels–Alderase antibodies 1E9 (11) and 39A11 (12). In the slower reacting substrate 1b, the H-bonding interaction would occur with the nonbonding electron pair of the bridge's NH group, which should rather slow down the retro-Diels–Alder reaction if one considers that the retro-Diels–Alder reaction does not occur in the protonated forms of the substrates (16, 17). Therefore, the participation of these hydrogen bonds could at most explain the 10-fold reactivity difference between the regioisomeric substrates 1a and 1b.

Conclusion

Shape complementarity to the transition state is a central paradigm that has guided catalytic antibody research over the last 15 years. To obtain a closer insight, we undertook a more quantitative analysis, which was made possible by the availability of x-ray structures of complexes with analogues of substrate, transition state, and product. The presence of a tryptophan residue at position H104 in retro-Diels–Alderases 10F11 and 27C5 instead of a tyrosine (9D9) correlates with increased catalytic potency. The larger aromatic residue clearly improves not only the absolute score of shape complementarity, but also increases the differential in Sc values between transition state and substrate, which is critical for explaining catalysis and not only substrate binding. It must be pointed out again that possible catalysis effects by orbital interactions involving residue H104 are inconsistent with existing data on other Diels–Alderase antibodies, and that the observed hydrogen bonds can only contribute to the observed regioselectivity. Thus, our x-ray structure analysis of retro-Diels–Alderase catalytic antibodies 10F11 and 9D9 suggests that aromatic residues can drive catalysis by acting as mediators of differential shape complementarity to a reaction's transition state over substrate or product.

Abbreviations

HOMO

highest occupied molecular orbital

LUMO

lowest unoccupied molecular orbital, Sc, shape complementarity index