Evidence for Antibody-Catalyzed Ozone Formation in Bacterial Killing and Inflammation

See allHide authors and affiliations

Science  13 Dec 2002:
Vol. 298, Issue 5601, pp. 2195-2199
DOI: 10.1126/science.1077642


Recently, we showed that antibodies catalyze the generation of hydrogen peroxide (H2O2) from singlet molecular oxygen (1O2*) and water. Here, we show that this process can lead to efficient killing of bacteria, regardless of the antigen specificity of the antibody. H2O2 production by antibodies alone was found to be not sufficient for bacterial killing. Our studies suggested that the antibody-catalyzed water-oxidation pathway produced an additional molecular species with a chemical signature similar to that of ozone. This species is also generated during the oxidative burst of activated human neutrophils and during inflammation. These observations suggest that alternative pathways may exist for biological killing of bacteria that are mediated by potent oxidants previously unknown to biology.

A central concept of immunology is that antibodies perform the sole function of marking antigens for destruction by effector systems such as complement and phagocytic cells (1). Work on antibody catalysis has demonstrated that the antibody molecule is capable of carrying out highly sophisticated chemistry, although there has been no direct evidence that this catalytic potential is used in nature (2). This view is consistent with the known organization of the humoral immune system, in that simple antigen binding is sufficient to activate more sophisticated effector systems and, thus, killing of pathogens can be achieved without the need to invoke any chemistry within the antibody molecule itself. Recently however, we found that all antibodies, regardless of source or antigenic specificity, can catalyze redox chemistry that is independent of antibody binding (3) and appears to be highly analogous to that carried out by the effector mechanism of phagocytic cells (4). When exposed to singlet molecular oxygen (1O2*), antibodies oxidize water to produce H2O2 via the postulated intermediacy of H2O3 (5). In the present study, we examined whether this pathway might play any role in immune protective function of antibodies against bacteria and in inflammation.

Initial bactericidal studies focused on the gram-negative bacteriaEscherichia coli (XL1-blue and O-112a,c) (6). Given the known bactericidal action of1O2* itself (7), these studies required a 1O2* generating system that would not, on its own, kill E. coli (8) but would activate the water-oxidation pathway of antibodies. Negligible bactericidal activity against the two E. coli serotypes (∼107 cells/ml) was observed when hematoporphyrin IX (HPIX, 40 μM), an efficient sensitizer of 3O2(9), was irradiated with white light (light flux 2.7 mW cm−2) for 1 hour in phosphate buffered saline (PBS, pH 7.4) at 4° ± 1°C. However, addition of antigen-specific or nonspecific monoclonal antibodies (20 μM) to this system resulted in killing of >95% of the bacteria (Fig. 1A) (6). This bactericidal action seems to be a general property of antibodies, in that regardless of origin or antigen specificity, all antibodies display this activity. This bactericidal activity was a function of antibody concentration (Fig. 1B), irradiation time (Fig. 1C), and HPIX concentration (at a given light flux) (Fig. 1D). These observations support the key role of both 1O2* and the water-oxidation pathway of antibodies in this bactericidal activity.

Figure 1

Killing of bacteria by antibodies. (A) Bar graph showing survival of E. coliXL1-blue and O112a,c strains [reported as percent recovered colony forming units (CFUs) at the start of the experiment (t = 0 min)]. Black and light gray bars correspond to the same experimental conditions, except that the light gray groups (2, 4, 6, 8, 10, and 12) were exposed to visible light for 60 min, whereas the black groups (1, 3, 5, 7, 9, and 11) were placed in the dark for 60 min. The bacterial cell density was ∼ 107 cells/ml. Each data point is reported as the mean ± S.E.M. (n = 6) of E. coli XL1-blue (groups 1 to 6) and O112a,c (groups 7 to 12) are as follows, in PBS at 4°C. Groups 1 and –2: XL1-blue cells. Groups 3 and 4: HPIX, XL1-blue cells . Groups 5 and 6: XL1-blue–specific monoclonal antibody (25D11, 20 μM), HPIX, XL1-blue cells. Groups 7 and 8: O112a,c cells. Groups 9 and 10: HPIX, O112a,c cells. Groups 11 and 12: O112a,c-specific monoclonal antibody (15404; 20 μM), HPIX, O112a,c cells. (B) Concentration effect of O112a,c-specific monoclonal antibody, 15404, on the survival of E. coli O112a,c. Each data point is reported as the mean value ± S.E.M (n = 3). The [15404] that corresponds to killing of 50% of the cells (EC50) = 81 ± 6 nM (C) Effect of irradiation time on the bactericidal action ofE. coli XL1-blue–specific murine monoclonal antibody 12B2. Graph of irradiation time (2.7 mW cm−2) versus survival of E. coli XL1-blue in the presence of HPIX and 12B2 (20 μM). Each data point is reported as the mean value ± S.E.M (n = 3). The time of irradiation that corresponds to killing of 50% of the cells = 30 ± 2 min (D) [HPIX]-dependence of E. coliXL1-blue-specific murine monoclonal antibody 25D11 bactericidal action. Graph of survival of E.coli XL1-blue versus exposure to a range of HPIX concentrations in PBS at 4°C under the following conditions. (▪), XL1-blue cells in the dark, 60 min. (▴), XL1-blue cells in white light. (▵), 25D11 (20 μM), XL1-blue cells in the dark, 60 min. (⧫), 25D11 (20 μM), XL1-blue cells in white light for 60 min. (E) Effect of catalase on the bactericidal action of antibodies against E. coli XL1-blue [reported as percent recovered colony forming units (CFUs) at the start of the experiment (t = 0 min)]. Each group was irradiated with white light for 60 min at 4°C. The bacterial cell density was ∼ 107 cells/ml in PBS The experimental groups (1 to 7) were treated were as follows. Group 1: E. coli XL1-blue and HPIX. Group 2: E. coli XL1-blue and nonspecific murine monoclonal antibody 84G3 (20 μM). Group 3: E. coliXL1-blue HPIX and 84G3 (20 μM). Group 4: E. coliXL1-blue HPIX, 84G3 (20 μM) and catalase (13mU/ml). Group 5:E. coli XL1-blue and specific rabbit polyclonal antibody (20 μM). Group 6: E. coli XL1-blue HPIX) and specific rabbit polyclonal antibody (20 μM). Group 7: E. coliXL1-blue, HPIX, specific rabbit polyclonal antibody (20 μM) and catalase (13mU/ml). Each point is reported as the mean value ± S.E.M. of multiple experiments (n = 6). **Denotes aP value < 0.01 relative to controls at the same time point. No bactericidal activity was observed in any of the dark controls (data not shown) (F) Concentration-dependent toxicity of H2O2 on the viability of E. coli XL1-blue and O112a,c serotypes. The vertical hatched line is the concentration of H2O2 expected to be generated by antibodies during a 60 min incubation using the conditions described in Fig. 1 and (18). The value of 35 ± 5 μM H2O2 is the mean value determined from 12 different monoclonal antibodies from at least duplicate measurements.

Given that H2O2 is the ultimate product of the antibody-catalyzed water-oxidation pathway (3,5), we reasoned that this was the principal killing agent observed in our assays. Consistent with this notion, catalase completely protected against the bactericidal activity of nonspecific antibodies (Fig. 1E) (10). However, quantification of H2O2 toxicity on the two E. colicell lines revealed that levels of H2O2generated by nonspecific antibodies (35 ± 5 μM) were between 1 and 4 orders of magnitude below that required to kill 50% of the bacteria (Fig. 1F) (6, 11).

Gold-labeled secondary antibodies and electron microscopy were used to correlate the morphological damage to sites on the bacterial cell wall where antibodies were bound (6). In the bactericidal pathway, there are clear stages in which oxidative damage leads to an increased permeability of the cell wall and plasma membrane to water where killing is associated with the production of holes in the bacterial cell wall at the sites of antigen-antibody union (Fig. 2). The observed morphologies induced by antibody-mediated killing are similar to those seen when bacteria are destroyed by phagocytic cells (12).

Figure 2

E. coli O112a,c cell after exposure to antigen-specific murine monoclonal IgG (15404, 20 μM), HPIX in PBS and visible light for 1 hour at 4°C (<5% viable). To visualize the sites of antibody attachment gold-labeled goat anti-mouse antibodies were added after completion of the bactericidal assay (6). The arrow points to a puncture in the cell and plasma membrane.

The finding that the toxicity of H2O2 toE. coli was below that generated by antibodies forced us to re-examine the implication of the experiments with catalase, discussed above. H2O2 could potentially react with some other chemical species also generated by the antibody to produce the bactericidal molecule(s). Thus, by destroying H2O2, catalase prevents the formation of these species. Alternatively, other species formed on the way to H2O2 may also be substrates for catalase. In the course of exploring which bactericidal agents might contribute to H2O2-mediated killing, we observed that one of the antibody-generated oxidants possesses the chemical signature of ozone (O3).

Theoretical calculations have shown that ozone is a plausible intermediate that could be produced during the water-oxidation pathway (13). Although ozone itself is highly bactericidal, there also exists a reaction between H2O2 and O3 that is termed the peroxone process. This process is exploited at the industrial scale for water purification, where it has been reported that a combination of H2O2 and O3 is far more toxic to bacteria than either alone (14, 15).

Under the aqueous conditions used in our assays, ozone is quite long lived [half-life (t 1/2) = 66 s] (6). Therefore, we began a search for O3 production during the water-oxidation pathway by antibodies using indigo carmine 1, a sensitive probe for O3 detection in aqueous systems (16,17). Conventional ozonolysis of 1 in aqueous solution leads to bleaching of the characteristic absorbance of1max = 610 nm; ɛ = 20,000 M−1cm−1) and the formation of isatin sulfonic acid 2 (Fig. 3A).

Figure 3

(A) Oxidation reaction of indigo carmine 1 to isatin sulfonic acid 2. (B) Progress of isatin sulfonic acid 2 production from indigo carmine 1 (1 mM) during UV irradiation (312 nm, 0.8 mW cm−2) of antibodies in PBS in the presence and absence of catalase (6). Each point is reported as the mean ± SEM. of at least duplicate determinations. Linear regression analysis was performed with Graphpad Prism v.3.0 software;v = rate of formation of 2. (▿), Sheep polyclonal IgG (20 μM) v = 34.8 ± 1.8 nM/min; (□); murine monoclonal antibody 33F12 (20 μM) v = 40.5 ± 1.5 nm/min; (•); sheep polyclonal IgG (20 μM) and soluble catalase (13 mU/ml) v = 33.5 ± 2.3 nM/min; (▪), murine monoclonal antibody 33F12 (20 μM) and soluble catalase (13 mU/ml) v = 41.8 ± 1.2 nM/min. (C to E) Electrospray ionization (negative polarity) mass spectra of isatin sulfonic acid 2[(M-H) 226, (M-H) 228 (18O), and (M-H) (2 x 18O)] produced during the oxidation of indigo carmine 1 (1 mM) in H2 18O (> 95% 18O) phosphate buffer (PB, 100 mM) at room temperature under conditions as follows: (C) Conventional ozonolysis (600 μM in PB) for 5 min. (D) Irradiation of HPIX with white light and sheep polyclonal IgG (20 μM) with white light for 4 hours. (E) Irradiation of HPIX with white light for 4 hours.

When antibody solutions in PBS were irradiated with ultraviolet (UV) light, under conditions where the water-oxidation pathway was functioning, in the presence of 1 (1 mM), catalase-independent bleaching of 1 and formation of2 was observed. (6) (Fig. 3B). The initial rate of antibody-mediated conversion of 1 into2 was linear, independent of the antibody preparation [sheep polyclonal immunoglobulin G (IgG) = 34.8 ± 1.8 nM min−1; 33F12 = 40.5 ± 1.5 nM min−1] and equivalent to 12% of H2O2 formation (3).

Though the oxidative cleavage of the C–C double bond of indigocarmine 1 is a sensitive probe for ozone detection, it is not a specific one. We have confirmed that1O2* also bleaches solutions of 1 to form 2 by oxidative double-bond cleavage (6). Given that 1O2* is generated by antibodies upon UV irradiation (3,5), we sought a means of analytically differentiating between oxidative cleavage of 1 to 2 by1O2* as opposed to one cleaved by O3. We observed that cleavage by O3 can be distinguished from cleavage by 1O2* through the different behavior of these two oxidation processes by observing18O incorporation from the reaction solvent H2 18O into the cleavage product 2(18). Experimentally, conventional ozonolysis of1 in phosphate buffer (10 mM, prepared with H2 18O) leads to the mass peak [M-H] 230 for 2 being obtained (Fig. 3D), a result of exchange of 18O of water into the amide carbonyl of 2 during the oxidation process. This mass fragment is not detected when 1 is oxidized by 1O2* (Fig. 3E) (19, 20). However, when sheep IgG (20 μM) and HPIX were irradiated with visible light (2.8 mW cm−2) in the presence of 1, oxidized product2 was formed that possesses the mass peak [M-H]230, suggesting that an oxidant with the chemical signature of ozone was among the reaction intermediates formed along the antibody-mediated water oxidation pathway (Fig. 3C).

Given the importance of the claim that ozone may be being generated during this process, we sought to further substantiate this observation with a chemical probe that contains a normal carbon-carbon double bond. The choice of the probes, 3- and 4-vinyl-benzoic acid (3 and4, respectively), was guided by their aqueous solubility coupled with ease of detecting the putative reaction products 3- and 4-carboxybenzaldehyde (5a, and 5b, respectively) and 3- and 4-oxiranylbenzoic acid (6a and 6b, respectively) (fig. S3) by high-performance liquid chromatography (HPLC) (6, 21).

When a solution of sheep polyclonal antibody (20 μM) in PBS was irradiated with UV light (312 nm, 0.8 mW cm−2), conditions where the water-oxidation pathway was activated in the presence of3 and 4 (1 mM) formed the oxidation products 3-carboxybenzaldehyde 5a and 3-oxiranyl benzoic acid6a (ratio 15:1, 1.5% conversion of 3after 3 h) and 4-carboxybenzaldehyde 5b and 4-oxiranyl-benzoic acid 6b (ratio of 10:1, 2% conversion to 4 after 3 hours), respectively. These products are also observed when 3 and 4 are ozonolized in PBS in a conventional way (6). In sharp contrast,1O2* generated by HPIX and visible light does not cause any detectable oxidation of either 3 or4 under these conditions. These observations with the orthogonal ozone probes 3 and 4 directly parallel the experimental observations with indigo carmine 1 and add further support to the notion that an oxidant with the chemical signature of O3 is generated during the antibody-catalyzed water-oxidation pathway.

Neutrophils are central to a host's defense against bacteria and have been shown to have antibodies on their cell surface (6,22) and to have the ability, upon activation, to generate a cocktail of powerful oxidants that have been suggested to include1O2* (23, 24). Thus, these cells may offer a nonphotochemical biological source of1O2* that, with antibodies, might be capable of processing this substrate via the water-oxidation pathway. Also, utilization of a cellular source of 1O2 by antibodies offers a broader potential biological context for physiological conditions that go beyond those that require light activation.

After activation with phorbol myristate (1 μg/ml), human neutrophils (1.5 × 10 7 cells/ml) produce an oxidant that cleaves indigo carmine 1 and generates isatin sulfonic acid2 (Fig. 4A) (25–27). Between 40 and 50% of the possible yield of isatin sulfonic acid 2 (25.1 ± 0.3 μM of a potential 60 μM) from 1 occurs during the neutrophil cascade, revealing that a high concentration of this oxidant is generated during the oxidative pathway. When this same experiment is carried out with nuetrophils in H2 18O water, between 50 and 75% of the amide carbonyl oxygen of 2incorporates 18O, as shown by the intensity of the [M-H] 230 mass peak in the mass spectrum (MS) of the isolated cleaved product 2 (Fig. 4B). This 18O incorporation with neutrophils parallels the observation with both conventional ozonolysis and antibody-mediated oxidative cleavage of1. Work must continue in order to determine whether the formation of this oxidant with the chemical signature of ozone by neutrophils is solely due to the antibody-mediated water oxidation pathway, or whether neutrophils are also capable of forming this oxidant when activated.

Figure 4

(A) Oxidation of 1 (30 μM) (▪) and formation of 2 (▴) by human neutrophils (PMNs) activated with phorbol myristate in PBS at 37°C. No oxidation of 1 occurs with PMNs that are not activated (data not shown). Neutrophils were prepared as previously described (25). (B) Negative-ion electrospray MS of the isatin sulfonic acid 2 produced during the oxidation of1 by activated human neutrophils, under the conditions described in (A).

To put these experiments into a physiological context, we studied an inflammatory model in vivo. The Arthus reaction is the classical inflammatory response that occurs when antigen-antibody union occurs in a tissue (28). In an immunologically competent host, antigen-antibody union activates complement, thereby initiating an inflammatory cascade, which includes the release of chemokines and granulocyte migration to the site where the immune complexes are formed.

A reversed passive Arthus reaction was generated in the skin of Sprague-Dawley rats using a bovine serum albumin (BSA)–antibody to BSA system (6). The inflammatory response at the site of antibody-antigen union peaked at approximately 8 hours and was associated with the classical gross and histological features of an Arthus reaction. Analytical studies were carried out on punch biopsies of either the inflammatory lesion or control skin samples from adjacent sites. These tissue samples were rapidly isolated and homogenized in a solution of indigo carmine (200 μM) in either PBS or PBS containing >95% H2 18O. Bleaching of the indigo carmine solution was observed in the inflammatory lesion but not with any control samples. The oxidation of 1 by the Arthus tissue was immediate and corresponded to ∼10% of the starting concentration (20 μM) of 1, as determined by UV absorbance at 610 nm. HPLC analysis confirmed that the oxidation of 1 was accompanied by formation of 2 as occurred in the experiments with purified antibodies and isolated neutrophils (Fig. 5A). MS analysis of 2 formed by the Arthus biopsies that had been placed into H2 18O-containing PBS revealed that the [M-H] 230 fragment was present, confirming that this inflammatory lesion contains an oxidant with the chemical signature of ozone (Fig. 5B).

Figure 5

Analytical evidence for the production of an oxidant with the chemical signature of ozone during the reversed passive Arthus reaction (6). A reversed passive Arthus was induced in Sprague-Dawley rats (150 to 200 g) by intravenous (i.v.) injection of BSA (100 μλ, 10 mg/ml) and a simultaneous intradermal injection of a polyclonal anti-BSA IgG (100 μl, 10 mg/ml). For control purposes, Sprague Dawley rats received an intradermal injection of a polyclonal non-BSA–specific IgG (100 μl, 10 mg/ml) preparation or PBS (100 μl) in adjacent skin sites. Eight hours after the i.v. administration of BSA (100 μλ, 10 mg/ml), the animal was then killed, and areas of injection were harvested using an 8-mm punch biopsy. Biopsy specimens from each reaction were added to a tissue homogenizer containing indigo carmine (50 μM). (A) HPLC traces of tissue homogenates generated from rat skin biopsies. The left trace is an analysis of the oxidation products of 1 generated from a control skin site that had been injected with PBS. The right trace was from an adjacent Arthus lesion generated by anti-BSA IgG. Note the presence of isatin sulfonic acid 2, retention time R T ∼ 9.2 to 9.3 min (B) Electrospray MS of the tissue homogenate obtained from an Arthus lesion, generated as detailed above, that had been added to a solution of indigo carmine (200 μM) in PBS containing > 95% H2 18O.

The results presented here may be biologically important in terms of the role of the immune system in both defense and its counterpart, induction of disease. The discovery of the bactericidal activity of antibodies in the presence of 1O2* is the first direct evidence that they can destroy their antigenic targets in the absence of complement or phagocytes. Further, our study demonstrates that a molecule, which appears to be ozone, is generated by antibodies during bacterial killing by activated neutrophils and in an inflammatory response in vivo. However, production of other short-lived trioxygen adducts that might have the same chemical signature as ozone cannot be discounted, although, as with ozone, evidence for the generation of such oxidants would be new to biology.

Ozone is a highly toxic, but short-lived substance. These properties would make it an ideal effector molecule, because any damage would be localized to the site of inflammation. Like other immune effectors, its synthesis in the systems studied here was triggered by antigen-antibody union that, in this case, occurs when activated neutrophils at the site of an infection generate 1O2*, which is the substrate for the antibody-catalyzed water-oxidation process. Ozone shares another key hallmark of immune effectors in that it not only kills but also functions as a signaling device that serves to amplify the inflammatory response by the production of nuclear factor kappa B (NF-κB), interleukin-6 (IL-6), and tumor necrosis factor–α (TNF-α) (29, 30). Lastly, the catalytic antibody field has shown that antibodies are capable of much more complex chemistry than simple binding. It has not been previously thought that this potential for complex chemistry plays a role in their in vivo function. However, in light of our data, one must now consider that all antibodies have an innate catalytic potential that may be exploited for host protection.

Supporting Online Material

Materials and Methods

Figs. S1 to S4


View Abstract

Navigate This Article