Antibody Catalysis of the Oxidation of Water

See allHide authors and affiliations

Science  07 Sep 2001:
Vol. 293, Issue 5536, pp. 1806-1811
DOI: 10.1126/science.1062722


Recently we reported that antibodies can generate hydrogen peroxide (H2O2) from singlet molecular oxygen (1O2*). We now show that this process is catalytic, and we identify the electron source for a quasi-unlimited generation of H2O2. Antibodies produce up to 500 mole equivalents of H2O2 from1O2*, without a reduction in rate, and we have excluded metals or Cl as the electron source. On the basis of isotope incorporation experiments and kinetic data, we propose that antibodies use H2O as an electron source, facilitating its addition to 1O2* to form H2O3 as the first intermediate in a reaction cascade that eventually leads to H2O2. X-ray crystallographic studies with xenon point to putative conserved oxygen binding sites within the antibody fold where this chemistry could be initiated. Our findings suggest a protective function of immunoglobulins against 1O2* and raise the question of whether the need to detoxify 1O2* has played a decisive role in the evolution of the immunoglobulin fold.

Antibodies, regardless of source or antigenic specificity, generate H2O2from 1O2*, thereby potentially aligning recognition and killing within the same molecule (1). Given the potential chemical and biological importance of this observation, the mechanistic basis of this process and its structural location within the antibody have been investigated. Together these studies reveal that antibodies, in contrast to other proteins, may catalyze an unprecedented set of chemical reactions between water and1O2*.

Long-term ultraviolet (UV) irradiation studies reveal that antibody-mediated H2O2 production is much more efficient than for non-immunoglobulin proteins (Fig. 1A). Typically antibodies exhibit linearity in H2O2 formation for up to 40 mole equivalents of H2O2 before the rate begins to decline asymptotically (Fig. 1B). Non-immunoglobulin proteins display a short burst of H2O2 production followed by quenching as photo-oxidation occurs (Fig. 1A). Also, antibodies can resume photoproduction of H2O2 at the same initial rate if H2O2 is removed by catalase (Fig. 1C). Thus, H2O2 reversibly inhibits its own formation. The apparent median inhibitory concentration (IC50) was estimated as 225 μM (Fig. 1E). Antibody-mediated photoproduction of H2O2 can also be saturated with molecular oxygen (apparent Michaelis-Menten constant for oxygen = 187 μM) (1), which, when allied with the H2O2 inhibition aspect, suggests a binding site process.

Figure 1

H2O2 production. (A) Production of H2O2 by immunoglobulins and non-immunoglobulin proteins. Assays were performed by near-UV irradiation (312 nm, 800 μW cm−2) of individual protein samples (100 μl, 6.7 μM) in PBS [10 mM sodium phosphate, 150 mM NaCl (pH 7.4)] in a sealed glass vial on a transilluminator (Fischer Biotech) under ambient aerobic conditions at 20°C. Aliquots (10 μl) were removed throughout the assay. H2O2 concentration was determined by the Amplex Red method (40, 41). Each data point is reported as the mean ± SEM of at least duplicate measurements: [•, human polyIgG; ○, horse polyIgG; □, sheep polyIgG; ▿, murine mIgG (WD1-6G6); ▵, human polyIgM; ◊, murine mIgG (92H2); ▪, β-galactosidase (β-gal); ▴, chick ovalbumin; ▾, α-lactalbumin; ⧫, bovine serum albumin]. (B) Long-term production of H2O2 by sheep polyIgG (6.7 μM, 200 μl). Near-UV irradiation for 8 hours in PBS in a sealed well of a 96-well quartz plate, H2O2 concentration was measured as described in (A). (C) A solution of murine mIgG PCP21H3 (6.7 μM, 200 μl) was irradiated in PBS in a sealed well of a 96-well quartz plate for 510 min. The H2O2was assayed by the Amplex Red assay and then destroyed by addition of catalase (10 mg, 288 mU) immobilized on Eupergit C. The catalase was removed by filtration, and the antibody solution was reirradiated for 420 min; rate (0 to 510 min) = 0.368 μM min−1(r 2 = 0.998); rate (511 to 930 min) = 0.398 μM min−1(r 2 = 0.987). This profile of continued linear production of H2O2 after catalase-mediated destruction of H2O2 is conserved for all antibodies assayed. (D) A solution of TCRαβ (6.7 μM, 200 μl) was irradiated as described in (C) for periods of 360, 367, and 389 min. The H2O2generated during each irradiation was assayed and destroyed as described in (C). The curvature in the progress curve above 30 mole equivalents conforms to the expected inhibition by H2O2 (see below); rate (361 to 727 min) = 0.427 μM min−1 (r 2 = 0.987); rate (728 to 1117 min) = 0.386 μM min−1(r 2 = 0.991). (E) Determination of IC50 of H2O2 on the photoproduction of H2O2 by horse polyIgG. A solution of horse IgG (6.7 μM) was incubated with varying concentrations of H2O2 (0 to 450 μM), and the initial rate of H2O2 formation was measured as described in (A). The graph is a plot of rate of H2O2 formation versus H2O2 concentration and reveals an IC50 of 225 μM.

Even after 10 cycles of UV irradiation followed by addition and removal of catalase (which generates ∼500 mole equivalents of H2O2), only a slight reduction (5%) is seen in the initial rate. Beside antibodies, the only other protein that we have found thus far to generate H2O2catalytically is the αβ T cell receptor (αβTCR) (Fig. 1D), which shares a similar arrangement of its immunoglobulin fold domains with antibodies (2). However, possession of this structural motif does not necessarily confer an H2O2-generating ability on proteins; β2-microglobulin, although a member of the immunoglobulin superfamily (3), does not generate H2O2.

The antibody structure is remarkably inert to the oxidizing effects of H2O2. SDS–polyacrylamide gel electrophoresis of antibody samples after UV irradiation under standard conditions for 8 hours revealed no significant fragmentation or agglomeration of the antibody. Also, the native and H2O2-treated structures of murine Fab 4C6 (4, 5) are superimposable at the level of side-chain positions, reinforcing the evidence of stability of the antibody fold in the presence of H2O2 (Fig. 2).

Figure 2

Superposition of the 4C6 combining site with and without H2O2 demonstrates that even the side-chain conformations within the binding site are preserved (light- and dark-colored side chains, pink for the light chains and blue for the heavy chains, correspond to + and – H2O2, respectively). Moreover, clear electron density for a benzoic acid ligand underscores that the binding properties of Fab 4C6 remain unaltered in 3 mM H2O2. The electron density map is a 2F obsF calc σ-weighted map contoured at 1.5σ. (Figure was generated with Bobscript.) Details of this structure will be presented elsewhere.

The photoactivity of the antibody appears to be driven through tryptophan (Trp) absorbance. An action spectrum of the antibody-mediated photoproduction of H2O2 and the corresponding absorbance spectrum of the antibody protein for wavelengths from 260 to 320 nm are virtually superimposable (Fig. 3). The maximal efficiency of H2O2production occurs at the same wavelength as the UV absorbance maxima of Trp in proteins (∼280 nm). We probed the efficiency of H2O2 production by horse immunoglobulin G (IgG) as a function of the efficiency of 1O2* formation via 3O2 sensitization with hematoporphyrin IX [quantum yield of singlet oxygen formation (φΔ) = 0.22 in phosphate buffer (pH 7.0) and visible light (6, 7)]. For every 275 ± 25 mole equivalents of 1O2* generated by sensitization, 1 mole equivalent of H2O2 was generated by the antibody molecule.

Figure 3

Absorbance spectrum (—) and action spectrum (–▪–) of horse polyIgG between 260 and 320 nm. The absorbance spectrum was measured on a diode array HP8452A spectrophotometer (maximum absorbance, 280 nm). The action spectrum was measured by placing an antibody solution [6.7 μM in PBS (pH 7.4)] in a quartz tube in the light beam produced by a xenon arc lamp and monochromator of an SLM spectrofluorimeter for 1 hour. H2O2 concentration was measured by the Amplex Red assay (40, 41).

The conversion of 1O2* to H2O2 requires two mole equivalents of electrons, and we have generated >500 equivalents of H2O2 per equivalent of antibody molecule with no notable reduction in rate. Thus, the ultimate electron source clearly cannot be the antibody itself. Both as an individual amino acid and as a constituent of proteins, Trp is particularly sensitive to near-UV irradiation (300 to 375 nm) under aerobic conditions, owing to its conversion to N′-formylkynurenine (NFK), which is a particularly effective near-UV (λmax = 320 nm) photosensitizer (8). However, photo-oxidation of Trp (the free amino acid) is accompanied by substoichiometric production of H2O2 (∼0.5 mole equivalents) during near-UV irradiation (Fig. 4A) (9), and the most efficient non-immunoglobulin protein at H2O2 photoproduction, β-galactosidase, generates only 7 mole equivalents of H2O2 from its 39 Trp residues (10) (Fig. 1A). Even if every photo-oxidizable residue (Trp, Tyr, Cys, Met, and His) were consumed, this could not account for 500 mole equivalents of H2O2 (7).

Figure 4

(A) Production of H2O2 by Trp (20 μM). The conditions and assay procedures were as described in Fig. 1A. (B) Effect of [Cl] on antibody-mediated photoproduction of H2O2. A solution of sheep polyIgG (▪, 6.7 μM, 200 μl) or horse polyIgG (▴, 6.7 μM, 200 μl) in PB (pH 7.4) was lyophilized to dryness and then dissolved in either deionized water or NaCl (aq.) such that the final [Cl] was 0 to 160 mM. The samples were then irradiated, in duplicate, in sealed glass vials on a transilluminator (800 μW cm−2) under ambient aerobic conditions at 20°C. Aliquots (10 μl) were removed throughout the assay; H2O2 concentration was determined by the Amplex Red assay (40, 41). The rate of H2O2 formation, ν, is plotted as the mean ± SEM versus [NaCl] for each antibody sample. (C) Effect of dialysis into EDTA-containing buffers on antibody-mediated photoproduction of H2O2. The photoproduction of H2O2 by two antibody preparations, mouse mIgG PCP21H3 and horse polyIgG, were compared before and after dialysis into PBS containing EDTA (20 mM). The conditions and assay procedures were as described in Fig. 1A. Each data point is reported as the mean ± SEM of at least duplicate measurements (•, murine mIgG PCP21H3 before dialysis; ▪, murine mIgG PCP21H3 after dialysis; ▴, horse polyIgG before dialysis; ⧫, horse polyIgG after dialysis).

The next most likely source is Cl, which is a suitable electron source for photoproduction of H2O2 via a triplet-excited state of an anthraquinone (11). We thus investigated the potential of Cl [present at 150 mM in phosphate-buffered saline (PBS)] as a reducing equivalent. However, the rate of H2O2 production by immunoglobulins was independent of [Cl] in the range 0 to 160 mM (Fig. 4B).

We also considered the possible role of metal ions. Although such ions could hardly be sufficiently abundant in antibodies to serve as an electron source, trace amounts of them might play a central role as catalytic redox centers. The following experiments allowed us to rule out the implication of trace metals in this process: (i) The rate of antibody-mediated photoproduction of H2O2 is unchanged before and after exhaustive dialysis of antibody samples with EDTA-containing buffer (Fig. 4C). (ii) After EDTA treatment of antibody samples, inductively coupled plasma–atomic emission spectroscopy (ICP-AES) reveals the presence of remaining trace metal ions in amounts far less than 1 part per million (7). (iii) For a trace metal to be implicated in this reaction, it must be common to all antibodies because all antibodies assayed have this intrinsic ability. It is generally accepted that metal binding is not an implicit feature of antibodies; this idea is consistent with our own analysis of antibody crystals as well as the ∼300 antibody structures available in the Brookhaven database.

All of our observations thus far pointed toward an electron source that does not deactivate the protein catalyst, could account for the high turnover numbers, and hence is quasi-unlimited. Our attention thus turned to a broader consideration of the chemical potential of1O2*.

The known chemistry of 1O2* (12) can be conceptualized as the chemistry of the superelectrophile “dioxa-ethene.” So we considered that a molecule of water may, in the presence of an antibody, add as a nucleophile to1O2* and form H2O3 as an intermediate. Water, in becoming oxidized to H2O2, would fulfill the role of the electron source.

Isotope experiments were undertaken to determine the source of oxygen found in the H2O2. Contents of16O/18O in H2O2were measured by modification of a standard H2O2 detection method: reduction with tris(carboxyethyl)phosphine (TCEP) (13) followed by mass-spectral analysis of the corresponding phosphine oxides (Fig. 5).

Figure 5

Electrospray ionization (negative polarity) mass spectra of TCEP [(M-H) 249] and its oxide [(M-H) 265 (16O) and (M-H) 267 (18O)] produced by oxidation with H2O2 [see (7) for assay conditions]. (A) After irradiation of sheep polyIgG (6.7 μM) under 16O2 aerobic conditions in H2 18O (98% 18O) PB. (B) After irradiation of sheep polyIgG (6.7 μM) under enriched18O2 (90% 18O) aerobic conditions in H2 16O PB (16). (C) After irradiation of sheep polyIgG under16O2 aerobic concentration in H2 16O PB. In this assay, H2 16O replaced H2 18O. (D) After irradiation of sheep polyIgG (6.7 μM) and H2 16O2 (200 μM) under anaerobic (degassed under argon) conditions in H2 18O PB for 8 hours at 20°C. (E) After irradiation of 3-methylindole (500 μM) under 16O2 aerobic conditions in H2 18O PB. Size-exclusion filtration was not performed because of the low molecular weight of 3-methylindole. TCEP was added to the 3-methylindole–containing PB solution. (F) After irradiation of β-gal (50 μM) under 16O2 aerobic conditions in H2 18O PB.

In the presence of oxygen, UV irradiation of antibodies leads to oxygen incorporation from water into H2O2(7). The relative abundance of the16O/18O ratio observed in the mass spectra of the phosphine oxide after irradiation of sheep polyclonal IgG (polyIgG) under conditions of saturating16O2 concentration in a solution of H2 18O (98% 18O) phosphate buffer (PB) was (2.2 ± 0.2):1 (Fig. 5A) (7). When the converse experiment was performed with an 18O-enriched molecular oxygen mixture (90% 18O) in H2 16O PB, the reverse ratio [1:(2.0 ± 0.2)] was observed (Fig. 5B) (14). These ratios exhibit good reproducibility (±10%, n = 10) (15) and were found for all antibodies studied (16).

The following control experiments were performed. First, under conditions of 16O2 and H2 16O, irradiation of horse polyIgG generates H2 16O2 (Fig. 5C). No incorporation of 18O occurs when H2 16O2 (400 μM in PB, pH 7.0) itself is irradiated for 4 hours in H2 18O. Thus, 18O incorporation into H2O2does not occur either by an acid-catalyzed exchange with water or by a mechanism that involves homolytic cleavage of H2 16O2 and recombination with H18O from water. To investigate the possibility that antibodies may catalyze both the production of H2 16O2 and its acid-catalyzed exchange with H2 18O, we determined the isotopic exchange of H2 16O2 (200 μM) in H2 18O (98% 18O) PB in the presence of sheep polyIgG (6.7 μM) after UV irradiation under an inert atmosphere. Only a trace of incorporation of 18O into H2 16O2 (<1%) was observed (Fig. 5D) (17).

The thermodynamic balance between reactants and products for the oxidation of H2O by 1O2* (heat of reaction ΔH r = +28.1 kcal/mol, Eq. 1a) (18) demands a stoichiometry in which more than one molecule of 1O2* must participate per molecule of oxidized water during its conversion into two molecules of H2O2. This stoichiometry assumes that no further light energy apart from that involved in the production of singlet from triplet oxygen is participating in the process. Qualitative chemical reasoning on hypothetical mechanistic pathways, together with thermodynamic considerations, makes the overall stoichiometries likely to be those shown in Eqs. 1b or 1c (heats of formation ΔH r° are reported in kcal/mol):Embedded Image(1a) Embedded Image Embedded Image(1b) Embedded Image(1c) Embedded ImageA recent report of a transition metal–catalyzed conversion of1O2 and water into H2O2via a tellurium-mediated redox process (19) provides experimental evidence for a process in which1O2* and H2O can be converted into H2O2. Hence, the energetic demands of this process can be overcome. At the heart of our considerations of a mechanism for the antibody-mediated photo-oxidation process is the hypothesis that addition of a water molecule to a molecule of1O2* forms H2O3 as the first intermediate on the way to H2O2. The antibody's function as a catalyst would have to be the supply of a specific molecular environment that would stabilize this critical intermediate relative to its reversible formation and/or would accelerate the consumption of the intermediate by channeling its conversion to H2O2. An essential feature of such an environment might consist of a special constellation of organized water molecules at an active site conditioned by an antibody-specific surrounding.

Although H2O3 has not yet been detected in biological systems, its chemistry in vivo has been a source of considerable speculation, and its in vitro properties have been the subject of numerous experimental and theoretical treatments (20–27). Koller and Plesnicar have shown that H2O3 reductively generated from ozone decomposes into H2O and 1O2* in a process catalyzed by a water molecule (26). Applying the principle of microscopic reversibility, we surmised that one or more molecules of water should also catalyze the reverse reaction. To delineate plausible reaction routes and energetics of such a process, we used first-principles quantum chemical (QC) methods [B3LYP (7)]:Embedded Image(2a) Embedded Image Embedded Image Embedded Image Embedded Image Embedded Image(2b) Embedded Image Embedded Image Embedded Image Embedded Image(2c)In these equations, all energetics are in kcal/mol. The direct reaction of water and 1O2* to give H2O3 is quite unfavorable, with an activation barrier of 64.7 kcal/mol (Eq. 2a). However, with the addition of a second or third water molecule, we find a concerted process that decreases the activation barrier to 31.2 kcal/mol and 12.0 kcal/mol, respectively. Indeed, these additional waters play a catalytic role (inEq. 2b, the H of the second water goes to the product HOOOH, simultaneous with the H of the first water replacing it). Note that the reverse reaction in Eqs. 2b and 2c has a barrier of only 19.2 kcal/mol or 0 kcal/mol, respectively, which suggests that H2O3 is not stable in bulk water or water-rich systems. Thus, we expect that the best site within the antibody structure for producing and using H2O3 would be one in which there are localized waters and water dimers next to hydrophobic regions without such waters.

We note that a 2.2:1 16O/18O incorporation ratio would coincide exactly with the value predicted for certain mechanisms in which two molecules of1O2* and two molecules of H2O are transformed into two molecules of H2O2 and one molecule of O2 (which would have to be3O2 for thermodynamic reasons). An example is a second-order nucleophilic substitution (SN2-type disproportionation) of two molecules of H2O3into H2O4 and H2O2, followed by the decomposition of the former into H2O2 and 3O2(28). Although our experimental evidence leads us to a hypothesis for the oxidation of water via H2O3, we have not discounted other mechanistic routes that may depend on a concert of events that are unique to antibodies.

Given the conserved ability of antibodies (regardless of origin or antigen specificity) and of the αβTCR to mediate this reaction, x-ray structural studies were instigated to search for a possible conserved reaction site within these immunoglobulin fold proteins. A key constraint for any potential locus is that molecular oxygen (either1O2* or 3O2 with a potential sensitizing residue, preferably Trp, in proximity) and water must be able to colocalize, and the transition states and intermediates along the pathway must be stabilized either within the site or in close proximity. Xenon gas was used as a heavy-atom tracer to locate cavities within the murine monoclonal antibody 4C6 (5, 7) that may be accessible to O2 (29–31). Three xenon sites (Xe1, Xe2, and Xe3) were identified (Fig. 6A), and all occupy hydrophobic cavities, as observed in other Xe-binding sites in proteins (32, 33). Superposition of the refined native and Xe-derivatized structures shows that, aside from addition of Xe, there is little discernible change in the protein backbone or side-chain conformation or in the location of bound water molecules.

Figure 6

The Xe binding sites in antibody 4C6 (7). (A) Standard side view of the Cα trace of Fab 4C6, with the light chain in pink and the heavy chain in blue. Three bound xenon atoms (green) are shown with the initial F obsF calc electron density map contoured at 5σ. (B) Overlay of Fab 4C6 and the 2C αβTCR (1TCR) around the conserved Xe1 site. The backbone Cα trace of VL (pink) and side chains (yellow) and the corresponding Vα of the 2C αβTCR (red and gold) are superimposed. (Figure was generated with Insight 2000.)

The Xe1 site is conserved in all of the antibodies we studied and the αβTCR (Fig. 6B). Xe1 is in the middle of a highly conserved region between the β sheets of VL (the variable region of immunoglobulin light chain), 7 Å from an invariant Trp. The Xe1 site is sandwiched between the two β sheets that constitute the immunoglobulin fold of the VL, ∼5 Å from the outside molecular surface. Xe2 sits at the base of the antigen binding pocket directly above several highly conserved residues that form the structurally conserved interface between the heavy and light chains of an antibody (Fig. 6A). The residues in the VLVH interface are primarily hydrophobic and include conserved aromatic side chains such as TrpH103.

The contacting side chains for Xe1 in Fab 4C6 are AlaL19, IleL21, LeuL73, and IleL75, which are highly conserved aliphatic side chains in all antibodies; only slight structural variation was observed in this region in all antibodies surveyed. Notably, several other highly conserved and invariant residues are in the immediate vicinity of this xenon site, including TrpL35, PheL62, TyrL86, LeuL104, and the disulfide bridge between CysL23 and CysL88. TrpL35stacks against the disulfide bridge and is only 7 Å from the xenon atom. In this structural context, TrpL35 may be a putative molecular oxygen sensitizer, because it is the closest Trp to Xe1. Comparison with the 2C αβTCR structure (3) and all available TCR sequences shows that this Xe1 hydrophobic pocket is also highly conserved in TCRs (Fig. 6B). Thus, the xenon experiments have identified at least one site that is both accessible to molecular oxygen and is in a conserved region (VL) in close proximity to an invariant Trp; an equivalent conserved site is also possible in the fold of VH (34). Analysis of the sequence and structure around these sites shows that they are highly conserved in both antibodies and TCRs. This finding may provide a possible understanding of why the Ig fold in antibodies and the TCR can be involved in this unusual chemistry (35).

As discussed previously (1), antibody-catalyzed production of H2O2 from1O2* may participate in antibody-mediated cell killing by event-related production of H2O2. Alternatively, antibodies may function in defending an organism against 1O2*. This postulate would require the further processing of H2O2 into water and 3O2by catalase (36). Because catalase is known to be an ancient protein arising as far back as archaebacteria (37), the question can be raised as to whether the structural element responsible for the catalytic destruction of1O2* is equally ancient and considerably precedes what we know today as antibodies. Singlet oxygen may even have played a decisive role in the initiation of the evolution of the immunoglobulin fold. Thus, it makes sense to search among ancient aerobic organisms for proteins that can accomplish similar chemistry.


View Abstract

Navigate This Article