115:412/508 Proteins & Enzymes spring 1997

Catalytic Antibodies

The basic ideas in the development of catalytic antibodies are simple, and indeed date back to a paper by Linus Pauling in American Scientist [1] in 1948, but could only have been brought to fruition recently, with the development of monoclonal antibodies; antibody libraries in E. coli, which are now coming on line, should make them even more available. The point of catalytic antibodies is that they are a way to generate an enzyme able to catalyze a specific reaction, without having to design the active site from the ground up and predict the entire protein structure necessary to fold to that active site.

The basic idea is that, if enzymes are described as stabilizing the transition state of a reaction by binding most tightly to it, having maximal interactions with that state, "one way to synthesize an enzyme is to prepare an antibody to a haptenic group which resembles the transition state of a given reaction. The combining sites of such antibodies should be complementary to the transition state and should cause an acceleration by forcing bound substrates to resemble the transition state." That is a direct quote from a book by Jencks in 1969 [2] . Since the immune system is considered able to make some 10¹⁰ different antibody molecules, it is likely to make a reasonable antibody to any molecule you can synthesize; the problems are to choose and synthesize the right transition state analog, and to find and make lots of the appropriate antibody. The second problem has been solved by the development of hybridoma technology for the production of monoclonal antibodies; I trust I don't have to explain this, which is summarized in a box on p. 28 of the Chemistry & Engineering News article [3] I am handing out. The development of the field, largely in the laboratories of Peter Schultz at Berkeley and of Tramontano and Lerner at the Scripps Clinic in La Jolla, California, has been to synthesize transition state analogs for more and more complex reactions, obtain antibodies to them, and demonstrate their catalytic activity. In 1993 Lerner was elected to the National Academy of Sciences.

The two groups burst into print simultaneously in Science [4] ^{,

[5]} in December 1986. Both these papers centered on the fact that phosphate esters, which are tetrahedral at the phosphorus atom, represent the tetrahedral transition state through which carboxylate esters are believed to pass in hydrolysis. Phosphoryl and phosphonyl esters have been useful transition state analog inhibitors in the study of the mechanism of thermolysin. Antibodies were elicited to the structure I have numbered 4 on p. 31 of the C&E News article. The dipicolinic acid group at the top was put in to generate a metal ion binding site; they didn't use it for that purpose, but only antibodies generated to this compound, not those to a phosphonate without this group, were catalytic. Initially they tried esters with a coumarin group, which I have drawn at the bottom of the page, as substrates. These made stable acyl enzymes. Then they tried a substrate with a p-acetamido group, shown just above 4 on p. 31. This was hydrolyzed; the half-time of hydrolysis of 5 µm substrate by 0.1 µm antibody was 16 min, vs. 4 min with a comparable level of hog liver esterase. A p-succinimidyl substrate was hydrolyzed, but not those with no p-substituent, only an acetamido group rather than trifluoracetamido on the left ring, or acetamido and trifluoracetamido groups interchanged. Hog liver esterase would hydrolyze these; the antibody is more specific than the enzyme. The succinimidylated substrate was more soluble, and saturation kinetics - approach to a V_max - could be demonstrated. K_m is 0.62 µm and k_cat80 sec^-1. The original hapten is a competitive inhibitor. They think the antibody has an imidazole at the active site, which acts as a general base catalyst in hydrolysis of the acetamido substrate, but forms an acyl enzyme by displacement of the more easily displacable coumarin group. My guess is that the idea is right, but that it is a lysine rather than a histidine, since acyl imidazoles are pretty reactive, should hydrolyze at a reasonable rate. The pH dependence was not observed; it could have provided evidence concerning this possible role - if the activity depended on the basic form of a group with pK_a around 7 it would indicate a histidine; a pK_a around 9 or 10 would be a lysine.

A later paper from Scripps [6] adds some details: a new antibody hydrolyzes the acetamido substrates, which have a lower spontaneous rate of hydrolysis; the rate of hydrolysis goes up with pH, but levels off with a pK_a= 8.9, and the protein is inactivated by nitration with tetranitromethane. The catalytic group is thus a tyrosine; presumably the phenolate anion acts as a general base catalyst.

The Schultz paper⁵ is actually more complete, since they worked with an already known and even sequenced antibody to p-nitrophenylphosphorylcholine, 6 on p. 31. This hydrolyzed the p-nitrophenyl choline carbonate just above it. The activity was specific for substrates with the choline group; showed saturation kinetics; was inhibited reversibly by the original hapten, and irreversibly by a reactive analog, p-diazoniumphenyl phosphoryl choline. At pH 7 the k_cat was 0.4 min^-1 and the K_m 208 µm, not nearly as good as the Scripps antibody. The log of V_max depended linearly on pH, which is to say that while this 'abzyme', as they are now termed, activates the substrate, it does not activate water. A real enzyme would use general base catalysis to convert an attacking H₂O molecule to OH^-. This is probably a role of the metal ion in thermolysin, and is suggested to be done by a histidine in the Scripps antibody. One could make an antibody to a transition state analog with an exchange-inert metal ion bound to it - certain transition metal ions, such as Cr⁺⁺⁺ and Co⁺⁺⁺, exchange ligands in their coordination shell only slowly and at elevated temperature, so that once complexes are formed with them they are stable. An antibody to such a complex would have a metal ion binding site built in, and might then bind a catalytic metal ion with exchange-labile coordinates.

They carried through on this idea [7] . They prepared a sort of tetrapeptide, which might be characterized as Phe-b-ala-phe-b-ala-gly, except that the N-terminal phe is not in peptide linkage, but shares the amino group with b-alanine (this was done by reacting the free amino group with phenylpyruvate and reducing with Na cyanoborohydride). See Fig. 9 on the next to last page. They cooked this with Co⁺⁺⁺(NH₂CH₂CH₂CH₂NH₂)₂(H₂O)₂ at 65° for 6 hr to get the Co⁺⁺⁺ complex with the phe COO^- and b-amino group. The complex was then attached to a protein to generate antibodies. The antibodies did bind a variety of metal-trien complexes, as shown by their ability to inhibit binding to the hapten, itself attached to BSA bound in a microtiter dish. The ability of the antibody plus metal-trien complex to hydrolyze phenylpropionyl-gly-phe-b-ala-gly was then tested, looking for the free amino group of the phe-b-ala-gly peptide product. This worked with, in decreasing order of relative activity, trien complexes of Zn⁺⁺, Ga⁺⁺⁺, In⁺⁺⁺, Fe⁺⁺⁺, Cu⁺⁺, Ni⁺⁺, Lu⁺⁺⁺, Mn⁺⁺ and Mg⁺⁺. The reaction is specific for this substrate and the phenylbutyryl homolog (14, 14a on the next to last page); phenylacetyl and formyl peptides don't work, nor does it work when b-alanine is substituted for glycine in the peptide. The activity is optimal around pH 6.5, though the pH optimum varies with the metal used. The turnover number is not great, 6x10^-4 s^-1. Cleavage is specifically at the gly-phe bond.

Benkovic, Lerner and co-workers reported [8] that antibodies to a p-nitrophenyl phosphonamide, 7 on p. 31 of the C&E News article, would hydrolyze the corresponding nitroanilide, just above it. They also showed hydrolysis of an unactivated benzyl ester, a-phenethyl (p-glutaramidyl)phenylacetate, by an antibody to the corresponding phosphonate [9] . In further studies [10] they showed that while ¹⁸O was incorporated into the carboxyl group of the product when the reaction was run in H₂¹⁸O, as one would expect, the abzyme did not catalyze exchange of ¹⁸O into the substrate, which you would expect it would if a tetrahedral intermediate were formed by addition of water to the carbonyl C; this does happen with the uncatalyzed reaction, i.e. if you just let substrate sit in H₂¹⁸O long enough (74 days at 37° C.) That such an exchange is not seen can mean one of several things: either the tetrahedral intermediate always proceeds to hydrolysis, never going back to substrate; or the two oxygen atoms in the intermediate are not equivalent, so that the added O is always lost when the intermediate goes back to substrate; or the tetrahedral intermediate is with a group on the enzyme, and is followed by an acyl-enzyme intermediate, as in chymotrypsin. The rate of the antibody-catalyzed reaction increases with pH, but then levels off, indicating a pK_aof about 9 in both the log V_max and log V_max/K_m profiles (that the pK_a is the same indicates that binding of substrate is at rapid equilibrium compared to further reaction, which is not surprising). The pH dependent rate, but not the pH-independent rate, shows a strong deuterium isotope effect, k_h/k_d = 3.8, suggesting that different rate-limiting steps are involved in the pH-dependent and independent rates. More than that they wouldn't speculate. I'll speculate: the rate-limiting step is attack of a hydroxyl, like the serine in chymotrypsin, to form the tetrahedral intermediate and thence an acyl enzyme; this attack is aided by general base catalysis at lower pH, but above pH 10 the hydroxyl is ionized anyway and general base catalysis is not needed.

Another group at Scripps [11] synthesized a hapten analogous to what is believed to be the transition state of the Claisen rearrangement of chorismate to prephenate, catalyzed by chorismate mutase; see structure at top right of p. 33. Only one of the 15 cloned antibodies they got had the enzymatic activity, but this one did catalyze the reaction 190-fold over the uncatalyzed rate, with a k_cat of 1.2 x 10^-3 s^-1 and a K_m 5.1 x 10^-5 m. The free hapten is a competitive inhibitor. They looked at the temperature dependence of the reaction: the entropy and enthalpy of activation are -22 cal/mol^.°K and 15 kcal/mol, vs. -12.85 cal/ mol^.°K and 20.71 kcal/mol for the real enzyme. The antibody works more by lowering the enthalpy of activation; the higher entropy of activation may indicate that it needs to change conformation during the reaction. A lot more has been done on this reaction, including X-ray crystallography of a catalytic antibody to locate the groups which lower the enthalpy of activation. See for instance Wiest & Houk, J. Am. Chem. Soc. 117:11628-11639 (1995).

Tramontano's group [12] has also made antibodies to fluorescein and used them to catalyze the reduction of resazurin to resorufin by sulfite. One of these antibodies had k_cat = 0.02 s^-1, K_m for resazurin 0.6 µm, K_m for sulfite 3 mm, 3x10⁵ x the uncatalyzed reaction rate. The K_d of resazurin from the abzyme is 3x10^-8 m, indicating that turnover is faster than dissociation from the binary complex! Not only fluorescein (K_i = 10^-10 m), but small ions such as F^-, Cl^-, Br^-, I^-, SO₄⁼, HPO₄⁼and benzoate inhibit. The pH dependence shows an acidic optimum and an inflection point around 6.7. Chemical modification of his, arg and tyr all reduce activity, the last two almost completely.

Another approach is use of antibody specificity to introduce a covalently attached reactive group into an antibody's binding site, which can then aid in action on a substrate. For instance, Pollack and Schultz [13] reacted MOPC 315, an antibody to dinitrophenyl groups, with dinitroanilides having a reactive aldehyde or bromoacetyl group linked to them by a disulfide bond. The reactive group presumably then reacts with a group on the protein outside the binding site. The disulfide bond was then reduced with dithiothreitol, leaving a sulfhydryl attached to the protein (after removal of the hapten, probably by dialysis). The substituted antibody then catalyzed hydrolysis of 7-hydroxylcoumaryl (3-dinitroanilido)propionate, k_cat = 1.45 x 10^-2 s^-1, K_m 1.2 µm, acceleration 60,000 fold compared to catalysis by the same molarity of dithiothreitol. A similar strategy was used to attach an imidazole group [14] by a disulfide bond to the sulfhydryl, although the catalysis wasn't as good as by the sulfhydryl, probably because the linkage to the imidazole was long and flexible, so that there would be a big entropy loss in its finding the substrate carbonyl.

A further innovation [15] is use of a catalytic antibody to form a product which would not be formed by the normal chemical reaction; in a sense it is catalyzing a new reaction. An epoxide oxygen can be displaced by a hydroxyl group several carbons away to form a five-membered ring (tetrahydrofuran) or six-membered ring (tetrahydropyran). Empirical observations codified in "Baldwin's rules" indicate that the five-membered ring will be favored in a chemical reaction, because it is easier for the attacking oxygen, the carbon site of attack, and the departing oxygen all to be in one line, and indeed it is the only product observed in the acid-catalyzed reaction (Fig. 1, first page of handout). Janda et al. made an analog which is the N-oxide of a substituted piperidine (N in a six-membered, saturated ring), as analog for the transition state en route to the tetrahydropyran product, attached it to keyhole limpet hemocyanin, and made monoclonal antibodies. Two did form the desired product with the six-membered ring, one very specifically, acting on only one enantiomer of the substrate and producing one product enantiomer. The K_m and k_cat were 356 µM and 0.91 min^-1 respectively. These results were further commented on in the same issue of Science by Sam Danishevsky [16] , who was responsible for the original idea of the in-line transition state.

I refer you to three other review articles [17] ^{,

[18]}^{,

[19]}^,which describe yet other antibody-catalyzed reactions. My conclusion in 1993 was that the potential of catalytic antibodies is in one sense great, in another sense not: having obtained a moderately effective catalytic antibody, there is no obvious way to improve it, as evolution has improved enzymes (except that you can make a better transition state analog, and look through more clones). If you get the antibody from an E. coli antibody library, you have it cloned and can do site-directed mutagenesis - when you know what changes you want to make - or try to mutate and select for better versions, if you have an appropriate screen. But if no enzyme exists to catalyze a reaction, it may be possible to make one, if a not very efficient one, by looking for the appropriate antibody. Even nature does this: an auto-antibody was found in man which accelerated the hydrolysis of a gln-met peptide bond in the vasoactive intestinal peptide! [20] .

I supplement this with a more recent paper, Transition-state stabilization as a measure of the efficiency of antibody catalysis [21] , J.D. Stewart & S.J. Benkovic, Nature 375:388-391 (1995)

They draw a thermodynamic cycle comparing the                                    ^Kn^†
rate of an uncatalyzed reaction SÆS^†ÆP with an anti-                       Ab+S € Ab + S^† Æ Ab + P
body-catalyzed reaction via the ground-state complex                       Ø≠K_m           Ø≠K_i
Ab·S (dissociation constant K_m) and the transition-                            Ab·S €   Ab·S^† Æ Ab + P
state complex Ab·S^† (whose dis-sociation complex is                                  K_Ab^†
assumed ≈K_i, the dissociation complex of the antibody complex with the original hapten which induced it. Then K_m/K_i ≈ k_cat/k_uncat (remember that K_m, K_i are dissociation constants, K_m should be bigger than K_i); K_m/K_i represents available binding energy for catalysis by stabilizing the transition state. For two-substrate reactions the relationship is (K_m1,K_m2/K_i) ≈ k_cat/k_uncat. The available data, plotted as log (k_cat/k_uncat) vs. log (K_m/K_i), generally support this (especially for the bimolecular reactions); see Fig. 2 on last page of the handout). Deviations are explained by 1) errors in determination of K_i resulting from using equations which assume [I]>>[Ab], 2) deviations between the hapten structure and the true transition state structure (e.g. phosphonic acids vs. carbon tetrahedral intermediates), 3) assumptions that there is only one important transition state (probably more valid for antibodies than for evolved enzymes). Also, the antibody might work by a different mechanism.

This describes understanding of favored reactions; can an antibody accelerate a previously unobserved reaction pathway, given enough K_m/K_i? It might either catalyze formation of a high-energy intermediate which in solution could yield many products, but direct its breakdown to a single product; or it might affect the relative heights of transition-state barriers between the ground state and various products. They give one example of the first type of process, which gives 98% of a cyclohexanol by ring closure of an olefin, and four different examples of the second, including the reaction (ref. 15) which yields a pyran rather than the energetically favored furan (see figures at bottom of first page of handout, and Fig. 3 on last page). Another example is formation of a peptide from acetylvaline p-nitrophenyl ester + tryptophanamide; the antibody blocks access of water to the transition state, rather than allowing hydrolysis of the nitrophenyl ester.

Free energy differences (∆∆G^‡) of 2-3 kcal/mol are sufficient to favor one pathway over another (yield ≥98%). Overall k_cat/k_uncat values of ≈10⁶ seem about the limit for abzymes (K_m/K_i = 10^-4M/10^-10M); the substrate binding has to be good enough for a reasonable rate at reasonable concentrations (mM). This points out the importance of K_i, whether for improving the rate of the reaction or improving its specificity. Much higher affinities for the transition state can be attained, enzymes have values as great as 10^-24 M. Enzymes can also destabilize the ground state (and intermediates before the transition state), although there is argument about this.

"In the case of enzymes, binding of transition-state analogues almost universally involves a slow isomerization of the initial encounter complex to the tight-binding form." (Ref. 19 on page 1 of the handout.) Abzymes however are derived merely for maximal binding to the transition-state analogue, the one case where binding has been carefully studied is a one-step reaction; they do not generally include the additional factor of structural rearrangement to improve transition-state binding. This may explain why some abzymes do not catalyze as well as expected from their K_m/K_i values, and suggests a line of research to make better abzymes (though how to do it is not easy; perhaps study X-ray structures of complex of transition-state inhibitor with a real enzyme?)