Flavobodies : generation and redox properties of flavinbinding antibodies

6.1. Introduction

Nature provides a pool of thousands of different enzymes. These highly evolved catalysts selectively accelerate chemical reactions. Natural biocatalysts, however, are often not stable under the more extreme conditions under which chemical processes are performed, such as high temperature or pH. Moreover, there are many reactions for which no natural enzyme is known. Therefore, the development of artificial enzymes has been given serious attention the last decade.

These artificial enzymes can be divided in three categories: synzymes, molecular imprinted polymers (MIPs) and catalytic antibodies (or abzymes). The first category comprises supramolecular systems such as cyclodextrines an porphyrins that are designed to form a substratebinding pocket containing catalytically active groups. The second category consists of polymers formed in the presence of a target molecule that resembles a transition state of a reaction and acts as a template for polymerization; upon removal of the template, a specific cavity is created. The third category, catalytic antibodies, has been extensively described in 1.3 of this thesis. Compared to natural enzymes, that accelerate reactions Up to 10 ¹⁷fold [1], artificial enzymes are slow. Catalytic antibodies are generally more efficient than synzymes or MIPs; the best catalytic antibodies give rate accelerations of 10 ⁶fold  whereas synzymes at their best give 10 ³-10 ⁴fold accelerations and MIPs are even slower with 10-100 fold rate enhancements [2,3]. Thus catalytic antibodies are at present the best artificial enzyme system, which is not surprising since catalytic antibodies use the same building blocks as enzymes (i.e. amino acids). These highly versatile building blocks differ in size, shape, charge, polarity and chemical reactivity and thus form a reservoir of chemical reactivity that can be applied to build tailored biocatalysts. Moreover, catalytic antibodies are, compared to the other artificial enzymes, relatively easy to generate and to modify.

For these reasons, we chose the catalytic antibody system to attain our goal: the development of an artificial enzyme mimicking a redox enzyme (discussed below).

6.2. Design and development of a redox-active catalytic antibody

To extend antibody catalysis to cofactor-dependent redox reactions, we chose to mimic flavoprotein oxidases. To achieve this goal the antibody must bind flavin ("flavobody") and thus be elicited against a flavin (analogue). We designed a flavin- based hapten, that served a threefold goal. Firstly, it should be in a reduced flavin conformation since in oxidases the flavin cofactor is reduced upon substrate oxidation. Secondly, it should also introduce a substrate binding pocket in the antigen binding site. Thirdly, it must introduce a hydrophobic pocket adjacent to N(5) or hydrogen bonds to the N(5) or the C4a position of the flavin, since this will facilitate reoxidation of the flavin in the antigen binding site. The hapten N(5)-benzoyl-N(10) -(ribityl succinimide) ester flavin seems to serve this threefold goal. The N(5) substituent both stabilizes the reduced conformation of the flavin (since 1,5-dihydro reduced flavin is not stable) and will induce a hydrophobic pocket adjacent to N(5) that could both serve as a substrate binding pocket and facilitate reoxidation of the flavin.

The antibody αRf _mod G ₁ generated against this hapten, using traditional hybridoma technology (see 1.4.1), stabilizes the reduced flavin conformation as was proven by time-resolved polarized fluorescence binding studies with oxidized and reduced flavin (Chapter 2). A +56 mV shift in redox potential compared to free flavin (ΔE ₀ ') for the two-electron reduction, was calculated from the dissociation constants for reduced and oxidized flavin (0.07 and 6 μM, respectively).  αRf _mod G ₁ accelerates the reoxidation of reduced flavin 12 times, which is only 10-100 times less than for most oxidases (chapter 5). We ascribed this difference with oxidases to the flavin conformation, i.e. planar versus bent. For example, in flavodoxins flavohydroquinone is forced into an energetically less favorable planar conformation which probably makes it a more powerful reducing agent. In most oxidases the flavohydroquinone is also held in a flat conformation. Antibodies, however, are adapted to the conformation the flavin possesses during antibody selection. Therefore the flavohydroquinone must be in the bent conformation when bound to  αRf _mod G ₁ .

In contrast, the scFv antibody fragment αFl _red 5 , elicited via an alternative strategy against (reduced) 1,5-dihydroflavin (Chapter 3), did not show an effect on the reoxidation rate of reduced flavin. This antibody causes a comparable shift in E ₀ '. Consequently, the rate acceleration must be due to the specific protein environment induced by the N(5) benzoyl substituent in the hapten.

Unfortunately,  αRf _mod G ₁   bound flavin could not be reduced by NADH, NAD(P)H or benzoyl formic acid. Therefore we must conclude that at present this flavobody cannot accomplish a catalytic redox cycle. On the one hand, screening more substrates might yield a compound that can reduce antibody bound flavin and so complete the cycle. On the other hand, it might be wiser to do a step backward and to redesign the hapten. For example, a C(5) substituted 5-deazaflavin will have the advantage of a flat conformation which will probably enhance the flavin reoxidation rate. Moreover, the N(5) substituent should be more carefully designed with a direct relation to the substrate to be used and preferentially possess some features of the transition state conformation of the substrate (for example a imine like N(5) substituent to mimic alanine oxidation).

6.3. Antibodies against an unstable hapten

Antibody technology has made a great stride forward since the development of hybridoma technology in 1975 [4]. The possibility to display functional antibody fragments at the tip of filamentous phage has allowed the development of antibody phage libraries. At present antibodies can be generated totally in vitro from naive (i.e. derived from an unimmunized source) phage display libraries [5]. With this technique animal cell culture and even mouse immunization are by-passed. Moreover, we demonstrated that antibodies can even be generated against the unstable hapten 1,5-dihydroflavin. This reduced flavin rapidly reoxidizes under aerobic conditions and can therefore not be used directly for the generation of antibodies in (the aerobic environment of) a mouse. However, using a naive antibody scFv phage display library, we could perform the selection under anaerobic and reducing conditions and were able to select an antibody, αFl _red 5  that specifically binds 1,5 dihydroreduced flavin (Chapter 3).

6.4. Regulation of the flavin redox potential by flavobodies

The role of the protein environment of the flavin cofactor in modulating its redox potential is still not thoroughly understood. Up till now the most comprehensive studies have been performed on flavodoxins, but other flavoproteins such as D- amino acid oxidase have been investigated too in this respect, although not as thoroughly [6]. In order to contribute to a better understanding of this issue, we have generated scFv antibody fragments against reduced (see chapter 3) and oxidized flavin (see chapter 4). These flavobodies are characterized by time-resolved and steady-state fluorescence spectroscopy and ELISA methods (mapping) and molecular modelling. The 3-dimensional models of the antigen binding sites were in good agreement with the experimental results. Binding of αFl _red 5  to flavin increases the redox potential, mainly due to an arginine residue interacting with the flavin N(1). This was in agreement with literature data on oxidases claiming that positive residues in the vicinity of ring C of the flavin (see fig. 1.2, chapter 1) increase the flavin redox potential. However, the flavobody αFl _ox is the most remarkable result of this approach: this flavin binding antibody, that does not resemble a natural flavoprotein, shows that when the electron deficient ring C of the flavin is not involved in binding, the redox potential is not significantly affected (chapter 4).

In conclusion, we like to state that flavobodies form tailored protein environments that can be selected from antibody phage display libraries against different flavin analogues, even if they are not stable under physiological conditions, within a period of a few weeks. For these reasons, antibody fragments are novel tools, additional to natural proteins, to study the effect of different protein environments on the flavin redox potential. Moreover, when expressed in E.coli, primary sequence information can easily be obtained and mutations can readily be performed which could give additional information on the effect of a specific amino acid on the flavin redox potential.

6.5. The future of redox-active flavobodies

We have shown that it is possible to generate antibodies that influence the flavin redox potential, even in a fashion comparable with natural flavoproteins (chapter 4). We have also shown that antibodies are capable of catalyzing the reoxidation of reduced flavin, like oxidases (Chapter 5). However, we failed to develop an antibody that catalyzes flavin reduction by substrate binding. The results described here taught us that the development of a redox active flavobody is an ambitious project. First of
all a better hapten should be designed (and synthesized) that will lead to an antibody that can bind both the flavin cofactor and the substrate in the correct conformation and orientation. Moreover, it should induce those amino acids that regulate the flavin redox potential in the desired way (based on current knowledge) and which activate the substrate when placed in the antigen binding site.

The second step is the generation of antibodies against such a hapten. This can be performed either in vivo or in vitro. In vitro strategies have the advantage that the hapten does not necessarily have to be stable under physiological conditions (Chapter 3). Moreover, different elution strategies can be applied in the panning procedure as described in Chapter 3. Here, our goal was to select antibodies specific for reduced flavin. In order to remove cross-reactive antibodies (i.e. antibodies that also bind oxidized flavin), we performed a pre-elution step with oxididized flavin prior to eluting the (specific) reduced flavin binding antibodies with reduced flavin. This strategy yielded an antibody population enriched for reduced flavin binding. This approach might be extended, for example, if one wants to select for antibodies binding ring C of the flavin. In that case pre-elution with a flavin analogue substituted on ring C (thus eluting ring A binders) might be useful. On the other hand, with in vitro selection of antibodies from naive antibody phage display libraries, the chance of success is dependent on the library size and diversity: large and diverse libraries give a higher chance of selecting antibodies with the desired properties (here we used a library with a repertoire size of 10 ⁸, at present, however, repertoires of>10 ¹⁰are available).

These improved selection strategies, however, are related to selection for binding and not for catalysis. For this much more powerful selection strategies are required. For enzymes a very elegant selection strategy, based on phage display (see 1.4.3), has been developed to select for catalytic activity [7]. In this approach enzymes displayed on phage are selected by their catalytic activity using a suicide inhibitor linked to a biotin. Active phage enzymes were covalently attached to the suicide inhibitor and could be captured with streptavidin-coated beads. For flavoenzymes suicide inhibitors have been reported [8] that inactivate D-amino acid oxidase via modification of an active site histidine. Using such an inhibitor, one can select for the presence of a histidine in the antigen binding site. If, based on such inhibitors, a selection system for redox active flavobodies could be developed, this would greatly enhance the chances of selecting a catalytic antibody. The best strategy would be first to select a large and diverse repertoire of phage antibodies against an optimally designed hapten, and then to randomly mutate these first generation flavobodies and select them for covalent binding to suicide inhibitors.

6.6. Conclusion: the future of catalytic antibodies in general

Although scientists have been quite successful in developing catalytic antibodies for a wide variety of chemical reactions, most catalytic antibodies are modest catalysts at best. This is not surprising since the evolution of enzymes has been going on for millions of years whereas catalytic antibodies are only in the first stage of their "evolution". During their evolution enzymes have been taught not only to bind selectively the transition states of their substrates but also to participate directly in bond making and breaking. Most catalytic antibodies only have been selected for binding a transition state analogue, although a few strategies for direct selection for catalysis. The principle of this approach is that a catalytic antibody is expressed in bacteria lacking an enzyme that is crucial to survive; only those bacteria that express a catalytic antibody that catalyzes the same reaction as the missing enzyme will survive. Improving catalysis by mutagenesis ("evolution") has been described (for example [9]).

One of the problems in selecting catalytic antibodies is to "fish" a weak antibody catalyst out of a pool of enzymes simultaneously produced by the hybridoma or bacterial cell. An enzyme contamination of less than 0.1% in the purified antibody sample can catalyze substrate conversion and insinuate that it is the antibody that is catalytically active. We have been working on a catalytic antibody, elicited against the hapten N(5)-benzoyl-N(10)-(ribityl succinimide) ester flavin, that seems to show phosphodiesterase activity to several carbohydrate-phosphate esters. In a way this is not surprising since these antibodies bind carbohydrates due to the carbohydrate substituent at the N(10) of the flavin hapten. However, we found that the hybridoma cells producing this antibody also produce several phosphodiesterases that could stick to the antibody during the purification steps and confuse the measurements for catalytic activity.

It remains to be seen if it is worth the effort to develop catalytic antibodies for reactions for which there are also enzyme equivalents. In this case the only benefit could be' that the process of designing and developing these antibodies teaches us more about how enzymes work [2]. However, there are other cases where catalytic antibodies, despite the efforts it costs to construct them, may be useful: for synthesis of compounds for which no enzyme alternative exists (for example [10,11]) and for prodrug activation used in therapy [12,13]. In the latter case the present situation is that enzymes are linked to antibodies recognizing tumor epitopes; the enzymes activate prodrugs for efficient tumor killing. The problem is that most enzymes are of non-human origin and therefore give rise to an immune response of the patient against the enzyme.

Humanized catalytic antibodies may replace these enzymes and thus circumvent this problem. In conclusion, there are practical applications for catalytic antibodies that are worth the efforts and costs to generate them. In other cases the scientific challenge is to learn more about enzymes by mimicking their evolution in an (relatively) extremely short period of time (i.e. decades).

6.7. References

1. Radzicka, A. and Wolfenden, R. A proficient enzyme. Science 267:90-93,1995.

2.Kirby, A.J. Enzyme mechanisms, models, and mimics. Angew.Chem.Int.Ed. 35: 707-724, 1996.

3. Mosbach, K. and Ramstrom, 0. The emerging technique of molecular imprinting and its future impact on biotechnology. Bio/Technology 14: 163-169, 1996.

4. Kohler, G. and Milstein, C. Continuous culture of fused cells secreting antibody of predefined specificity. Nature 256: 495-497, 1975.

5. Winter, G., Griffiths, A.D., Hawkins, R.E. and Hoogenboom, H.R. Making antibodies by phage display technology. Annu.Rev.Immunol. 12: 433-455, 1994.

6. Stankovich, M.T. Redox properties of flavins and flavoproteins. In: Chemistry and Biochemistry of Flavoenzymes, volume 1, CRC Press, Boca Raton, p. 401-425, 1990.

7. Soumillion, P., Jespers, L., Bouchet, M., Marchand-Brynaert, J., Winter, G. and Fastrez, J. Selection of beta.gif-lactamase* on filamentous bacteriophage by catalytic activity. J.Mol.Biol. 237: 415-422, 1994.

8. Walsh, C. Suicide substrates: mechanism-based enzyme inactivators. Tetrahedron 7: 871- 909, 1982.

9. Smiley, J.A. and Benkovic, S.J. Selection of catalytic antibodies for a biosynthetic reaction from a combinatorial cDNA library by complementation of an auxotrophic Escherichia coli: antibodies for orotate decarboxylation. Proc. Natl.Acad.Sci. U.S.A. 91: 8319-8382, 1994.

10. Reymond, J.L., Reber, J.L. and Lerner, R.A. Enantioselective, multi-gram scale synthesis with a catalytic antibody. Angew.Chem.int.Ed. 33: 475-477, 1994.

11. Gouverneur, V.E., Houk, K.N., Pascual-Teresa, B., Beno, B., Janda, K.D. and Lerner, R.A. Control of the exo and endo pathways of the Diels-Alder reaction by antibody catalysis. Science 262: 204-208, 1993.

12. Miyashita, H., Karaki, Y., Kikuchi, M. and Fujii, 1. Prodrug activation via catalytic antibodies. Proc.Natl.Acad.Sci.U.S.A. 90: 5337-5340, 1993.

13. Campbell, D.A., Gong, B., Kochersperger, L.M., Yonkovich, S., Gallop, M.A. and Schultz P.G. Antibody-catalyzed prodrug activation. J.Am.Chem.Soc. 116: 2160-2166, 1994.