A Quantitative Study of NF-κB Activation by H₂O₂: Relevance in Inflammation and Synergy with TNF-α¹

Reactive oxygen species (ROS)³ and in particular H₂O₂ have a key role in the protection against invading pathogens during the innate immune response. Upon invasion by a pathogen, NADPH oxidase in neutrophils and monocytes is activated, liberating ROS, which combined with other factors attack the pathogen (1). A deficient H₂O₂ production due to mutations in NADPH oxidase, as observed in patients with chronic granulomatous disease, increases the susceptibility to pathogens that do not release H₂O₂ (2) and deregulates the inflammatory response (3).

NF-κB is a key regulator of the immune system because its activation induces several genes related to the innate immune and inflammatory responses. The recognition of pathogens by the TLR family of membrane receptors, mediation of neutrophil adhesion to the endothelium and transmigration from blood vessels to tissue interstitium, production of the proinflammatory cytokine TNF-α, and its actions in target cells are all processes under NF-κB control (4). Accordingly, bacteria that block NF-κB activation disrupt the innate immune response (5, 6). Originally, it was proposed that NF-κB activation induced by diverse stimuli shared a common secondary messenger, the oxidant H₂O₂ (7). For example, LPS, a component of the outer membrane of Gram-negative bacteria, binds to TLR4, which activates NADPH oxidase, releasing H₂O₂ and activating NF-κB (8). However, nowadays the role of H₂O₂ in NF-κB activation in vivo is highly controversial. In fact, H₂O₂ does not activate NF-κB in many cell lines (9, 10). Inhibition of TNF-α-induced NF-κB activation by the antioxidants N-acetyl-l-cysteine and pyrrolidine dithiocarbamate, previously interpreted as supporting a role played by H₂O₂ (11, 12), was recently shown to be independent of their antioxidant function (13). H₂O₂ concentrations required to activate NF-κB are typically in the range 0.1 to 1 mM, which is much higher than the 5–15 μM range observed during inflammation (14, 15), where H₂O₂ is produced at high rates. Because during inflammation, cells are subjected simultaneously to both cytokines and H₂O₂, the combined actions of these species have been investigated in a number of studies. In two commonly used cell lines, Jurkat T cells and HeLa, H₂O₂ was found to either stimulate (16, 17) or inhibit (18, 19) NF-κB activation by TNF-α. Contradictory results were also obtained in closely related alveolar epithelial cell lines (20, 21, 22). The stimulatory effects have been explained by an independent pathway induced by H₂O₂ involving the small G protein Ras (20) or through phosphorylation of serine residues at the IκB kinase complex subunits (16), whereas the inhibitory effects were attributed either to direct inhibition of IκB kinase complex activity through cysteine oxidation (21) or to inhibition of the proteasome thus blocking IκB-α degradation (22). Thus, despite the wealth of data on the activation of NF-κB by H₂O₂, the fundamental question whether H₂O₂ will activate or inhibit NF-κB remains unanswered.

NF-κB activation has a dual and opposite dependence on oxidative events, because its translocation is favored by oxidative events in the cytosol while binding to DNA requires a reductive environment in the nucleus (23, 24, 25). So, higher levels of oxidative exposure can turn a potential positive stimulus by H₂O₂ into an inhibitory effect. We hypothesized that the contradictory results reported in the literature are due to the uncontrolled manner by which H₂O₂ was delivered to cells thus exposing cells to different experimental H₂O₂ concentrations. So, the aims of the present work are to address whether H₂O₂, at concentrations approaching those found in vivo, activates NF-κB, stimulates or inhibits the activation of NF-κB by TNF-α, and fulfils a regulatory role during inflammation. For the first time, NF-κB activation was investigated using a calibrated method that delivers H₂O₂ continuously, mimicking the endogenous production of H₂O₂. This method allowed a rigorous control of the extracellular H₂O₂ concentrations applied (in the range of 5 to 25 μM) as well as the extent of the exposure to H₂O₂ (from 15 min up to 6 h) (26). The studies were conducted in two epithelial cell lines, MCF-7 and HeLa cells, by measuring both the translocation of the p65 subunit of the NF-κB complex into the nucleus and the IκB-α levels in the cytosol. To further address the relevance of the activation of NF-κB by H₂O₂, microarray analysis of a broad range of genes regulated by NF-κB was performed in the presence of H₂O₂.

MCF-7 (European Collection of Cell Cultures) and HeLa cells (American Type Culture Collection) were grown in RPMI 1640 medium supplemented with 10% of FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM l-glutamine (all obtained from Cambrex). The MCF-7 cell line is immunologically active (27), undergoes NF-κB activation by H₂O₂ (28), and has been used as a model of the mammary epithelium (29). Despite being a breast cancer cell line it belongs to a less aggressive and noninvasive/metastatic group (30). HeLa cells are also immunologically active (31). DMSO, catalase (bovine liver), glucose oxidase (Aspergillus niger), IGEPAL CA-630, MTT, TNF-α (human recombinant), benzamidine, leupeptin, pepstatin, PMSF, and DTT were obtained from Sigma-Aldrich, Inc., Saint Louis, MO. H₂O₂ was obtained from Merck & Co. All polyclonal Abs were obtained from Santa Cruz Biotechnology.

MCF-7 and HeLa cells were plated onto 100-mm dishes 48 h before the experiment to achieve ∼1.8 × 10⁶ cells and 1.5 × 10⁶ cells per dish, unless otherwise referred. Fresh medium was added to cells 1 h before incubations. H₂O₂ concentration in the medium was determined on aliquots of the medium by measuring the oxygen released after catalase addition using an oxygen electrode (Hansatech Instruments). As described by Antunes and Cadenas (26), the H₂O₂ steady-state [H₂O₂]_ss was achieved by adding an initial dose of H₂O₂ at the desired concentration; simultaneously, and to keep H₂O₂ concentration constant (steady state) during the entire assay, an adequate amount of glucose oxidase was added to compensate for the rapid consumption of the initial H₂O₂ concentration by cells. The H₂O₂ concentration was checked periodically. Cell viability was assessed by the ability of cells to reduce MTT (32).

Intracellular concentrations of H₂O₂ were estimated from the determination of the H₂O₂ gradient across the plasma membrane as described by Antunes and Cadenas (33). This gradient (1/R), i.e., the ratio between H₂O₂ concentration outside and inside (cytosol) the cell, may be inferred from H₂O₂ consumption by intact cells over the sum of enzyme activities (mainly catalase and glutathione peroxidase) that consume H₂O₂ in disrupted cells.

H₂O₂ consumption by intact cells was determined by adding an initial dose of 90 μM H₂O₂ and following the decay of H₂O₂ concentration along time using an oxygen electrode. H₂O₂ consumption is reported as a first-order rate constant (k_intactcell). Catalase activity (reported as a first-order rate constant, k_catalase) was measured as previously described (34) using ∼3 × 10⁵ cells and in the presence of 1 μg/ml digitonin to permeabilize the plasma membrane. GPx activity (k_GPx) was measured as previously described (33, 35) in a cell extract (corresponding to 8 × 10⁵ cells) collected after centrifugation (5000 × g for 5 min) of a cell lysate obtained with 0.1% (v/v) Triton X-100.

Cells were washed twice with cold PBS, pH 7.2 (1.5 mM KH₂PO₄, 155 mM NaCl, and 2.7 mM Na₂HPO₄) and lysed with 500 μl of cytosolic lysis buffer (50 mM HEPES (pH 7.2), 2 mM EDTA, 10 mM NaCl, and 250 mM sucrose plus freshly added 2 mM DTT, 0.1% IGEPAL CA-630 (v/v), and protease inhibitors: 1 mM PMSF, 1.5 μg/ml benzamidine, 10 μg/ml leupeptin, and 1 μg/ml pepstatin). Cells were scraped up and transferred to a microcentrifuge tube, and the remaining cells were collected from the dish with 100 μl of cytosolic lysis buffer. Cytosolic proteins were collected after centrifugation at 3000 × g for 4 min at 4°C. The pellet was washed once with the cytosolic lysis buffer and then ressuspended with the nuclear lysis buffer (same as the cytosolic buffer except that 400 mM NaCl and 20% glycerol (v/v) were used, but neither sucrose nor IGEPAL CA-630 were added to the buffer). Cells were put on ice during 20 min and mixed by vortexing three times during this incubation. Nuclear proteins were collected after centrifugation at 10,000 × g for 10 min at 4°C. Protein concentration was assayed using the Bradford method (36).

Protein samples were separated by SDS-PAGE on an 8% polyacrylamide gel, followed by a semidry electroblotting of proteins onto a nitrocellulose membrane (Protan; Schleicher & Schuell). The membrane was blocked with 5% nonfat dry milk in PBS, followed by immunoblotting using primary Abs to p65 (sc-109; 1:400) and IκB-α (sc-371; 1:800), and a secondary Ab conjugated to HRP (sc-2004; 1:1000). All incubations were performed at room temperature, during 2 h for the primary Abs and 1 h for the secondary Ab. Signals were developed using the ECL chemiluminescence system (Amersham Biosciences). Immunoblot films were digitalized and analyzed with the ImageJ software (37). Control of protein loading was performed by analysis of the membrane stained with Ponceau S red. To quantify the percentage of increase or decrease of the p65 levels in the presence of both TNF-α and H₂O₂ relative to the sum of the individual effects, Equation 2 was used:

When the effect observed is equal to the sum of the individual effects (addictive effect), there is no modulation, and Equation 2 equals 1. Positive modulations (synergism) and negative modulations (antagonism) will be above or under 1, respectively. The denominator was subtracted from 1 because nuclear translocation of p65 has relative units. For example, if both individual activations are the same as the control, the subtraction from 1 is needed in order for the sum to be also 1.

RNA from treated cells was extracted using the Array Grade Total RNA isolation kit (GA-013) and ∼3 μg of RNA was used in the experiments. Labeling was performed overnight with biotin-16-uridine-5′-triphosphate (Roche) using the TrueLabeling-AMP 2.0 kit (GA-030). The probe was purified using the ArrayGrade cRNA cleanup kit (GA-012). Hybridization of 3 μg of probe to the oligo GEarray human NF-κB signaling pathway Microarray (OHS-025) was done overnight and detection was assayed by chemiluminescence using the detection kit. Signal intensity was quantified using the ImageJ software (37). All kits were obtained from SuperArray Bioscience Corporation, and the protocols were followed according to manufacturer instructions. Gene expression induced by both H₂O₂ and TNF-α ((H + T)_i) was estimated as the average of the ratios between the expression caused by the simultaneous addition of the two agents divided by the expression caused by TNF-α (T_i) in each experiment, multiplied by the average expression caused by TNF-α stimulation in all experiments (T_average):

Results are presented as the average ± SD. Results were analyzed by using the two-tailed one sample t test comparing the average with 1: ∗∗∗, p < 0.001; ∗∗, p < 0.01; ∗, p < 0.05; •, p < 0.10.

In this study, delivery of H₂O₂ was performed using the steady-state titration method, in which cells are exposed to known and constant H₂O₂ concentrations ([H₂O₂]_ss). To show the advantages of this method, we compared it with two other approaches: 1) the bolus addition, by far the most common method, in which H₂O₂ is added at the beginning of the experiment; and 2) the glucose oxidase method, in which glucose oxidase is added to continuously produce H₂O₂ from the oxidation of the glucose present in the medium. As shown in Fig. 1,a, an initial bolus addition of 1 mM H₂O₂ was consumed by cells, and a very low concentration of H₂O₂ was measured after 2 h. Therefore, the bolus addition is clearly unsuitable to investigate processes where H₂O₂ is present for more than a couple of hours, such as the inflammation process. Concerning shorter periods, this method is also inadequate because high initial doses of H₂O₂ are needed to observe the response under investigation due to the fast consumption of H₂O₂ by the cells. Thus, the high initial dose possibly leads to disruption of cellular homeostasis, and the observed effects may be artifactual. When cells were exposed to glucose oxidase there was a gradual increase in H₂O₂ concentration that culminated in a near steady state after ∼2 h (Fig. 1,b). So, when using this approach it is important to note that the desired H₂O₂ concentration is not obtained immediately, which becomes a problem when short exposures are required. Concerning longer periods of exposure, the glucose oxidase method could a priori be considered adequate. However, the simple exposure of cells with glucose oxidase (or to a bolus addition of H₂O₂) represents an uncalibrated and uncontrolled way of delivering H₂O₂ to cells. To illustrate this point, cells were exposed to H₂O₂ at two different cell densities (Fig. 1, a and b). As can be seen the H₂O₂ concentration profile, including the near steady state reached after 2 h with glucose oxidase, depends on the cell density used. Other factors, such as the cell type (different cells consume H₂O₂ with different rates), or the incubation medium (different medium can consume (unpublished observation) or produce (38, 39) H₂O₂ with different rates), can potentially affect the actual H₂O₂ concentration reached in the assay.

In the steady-state titration, the desired [H₂O₂]_ss is given initially, together with the appropriate units of glucose oxidase that will produce H₂O₂ to the same extend of its consumption in the specific conditions of the assay, which include the cell density and the small day to day differences in cell behavior. As shown in Fig. 1 c, it was possible to maintain the desired [H₂O₂]_ss during the entire assay for the two cell densities. The key feature of this method is that it is calibrated on a daily basis because the amount of glucose oxidase added will be adjusted to match the consumption of H₂O₂ in the specific experimental conditions of the assay.

In conclusion, without measuring the H₂O₂ concentration achieved during the experiment it will not be possible to predict the concentration of H₂O₂ that cells are subjected to and the comparison of data obtained in different laboratories is not possible. The same initial dose of H₂O₂ or of glucose oxidase can represent very different levels of stress, even if the same line is used, and cause contradictory observations. In the steady-state titration, this problem is overcome by measuring the H₂O₂ concentrations in every assay and by adjusting the amount of glucose oxidase to reach the desired H₂O₂ concentration. This may be particularly relevant for biological processes subjected to a complex control by H₂O₂ levels, such as the dual dependence of NF-κB activation on H₂O₂ (23, 24, 25).

H₂O₂ diffusion across biomembranes is not “free” and gradients are formed when the source of H₂O₂ is separated from the sink by a biomembrane (33, 40). Because H₂O₂ was added extracellularly and NF-κB activation occurs in the cytosol, it is important to estimate the actual cytosolic H₂O₂ concentrations achieved in our experiments. An H₂O₂ gradient across the plasma membrane of ∼2 was obtained by applying Equation 1 (Materials and Methods and Table I), an equation that was previously applied by us in yeast and in Jurkat T cells (33, 40). Therefore, for the extracellular H₂O₂ concentrations in the 5 to 25 μM range used in this work, the corresponding cytosolic concentrations in MCF-7 cells were in the 2.5 to 12.5 μM range.

To investigate the relevance of H₂O₂ as an NF-κB activator, p65 translocation into the nucleus was measured by western blot. We started by applying extracellular [H₂O₂]_ss in the range between 5 and 25 μM, which are close to the H₂O₂ extracellular levels during inflammation (14, 15). In these conditions, cell viability was unaffected (not shown) and p65 translocation was induced significantly, up to 4-fold by H₂O₂ (Fig. 2, a and b), with a slow monotonous nonoscillatory kinetics. Removal of H₂O₂ reversed NF-κB activation, and the simultaneous addition of [H₂O₂]_ss and catalase eliminated NF-κB activation after 4 h thus showing that the reversible NF-κB activation is caused by H₂O₂ production (Fig. 2,d). To assess whether this level of activation is quantitatively important, we compared it with the activation of NF-κB induced by TNF-α, a well-known NF-κB inducer. TNF-α concentrations, both in the physiological range (0.2 to 1.5 ng/ml (41)) and in the supraphysiological range (higher than 1.5 ng/ml, but often used in experiments (16, 23, 42, 43)) were used without affecting cell viability (not shown). For the physiological range, there was an up to 8-fold induction of p65 translocation, whereas for the supraphysiological range a 12-fold induction was observed (Fig. 2,c). The recently discovered oscillatory dynamics of NF-κB activation (Fig. 2 e) were also observed (44).

The differences observed on the levels of p65 translocation may reflect different mechanisms of NF-κB activation. Contrary to TNF-α, NF-κB activation by [H₂O₂]_ss was not associated with a decrease in IκB-α levels (Fig. 2, a–c) and was not affected by proteasome inhibition (not shown). Thus, H₂O₂ may either induce the phosphorylation of tyrosine 42 of IκB-α, which releases IκB-α from the complex with NF-κB without the subsequent degradation, allowing NF-κB translocation (45), or inhibit NF-κB efflux from the nucleus, allowing NF-κB to slowly accumulate in the nucleus due to the low basal NF-κB activation (46).

From these results it can also be extrapolated that NF-κB activation by H₂O₂ does not occur during normal metabolism, where intracellular H₂O₂ levels are in the 0.01–0.1 μM range (47), and that H₂O₂ is at best a poor NF-κB activator when compared with TNF-α in MCF-7 cells (Fig. 2, b and c). This comparison has physiological relevance because the extracellular [H₂O₂]_ss applied (5–25 μM) are probably only found in vivo when other inducers of NF-κB (e.g., TNF-α) are also present, such as during the inflammatory response.

Due to the weak capacity of H₂O₂ to activate NF-κB, and taking into account that during inflammation both H₂O₂ and TNF-α are formed, we investigated whether H₂O₂ could modulate NF-κB activation induced by TNF-α. Cells were exposed to either a [H₂O₂]_ss of 25 μM or TNF-α (0.37 ng/ml) alone, or to both agents simultaneously. In both cell lines, the effect of the simultaneous addition of TNF-α and H₂O₂ on p65 levels in the nucleus was significantly higher than the sum of the individual effects of TNF-α and H₂O₂ (Fig. 3, a and c), i.e., H₂O₂ had a significant positive modulatory effect—a synergism—on the activation of NF-κB by TNF-α. In MCF-7 cells, a synergism of 20% was first observed at 30 min, a time corresponding to a peak of TNF-α-induced NF-κB activation but to no activation of NF-κB by H₂O₂, while a maximal synergism of 40% was observed at 2 h (Fig. 3,b). In HeLa cells a synergism on NF-κB activation of ∼100% (Fig. 3,d) was observed at 2 and 4 h. This high synergism is easily visualized because NF-κB activation by [H₂O₂]_ss was almost negligible (Fig. 3 c).

Most published studies (21, 22, 43) report that H₂O₂ inhibits the activation of NF-κB by TNF-α, although a few report a positive effect (16, 20) in agreement with our results. We hypothesized that the inhibitory effects observed in most studies could be related to the delivery of H₂O₂ as a large bolus initial dose, which causes severe oxidation and thus inhibits NF-κB activation. As shown in Fig. 3, e and f, for MCF-7 cells and in Fig. 3, g and h, for HeLa cells, under the artifactual conditions of a bolus addition, TNF-α-dependent translocation of p65 is significantly inhibited by H₂O₂ (between 40 and 60%). This negative modulation decreases when H₂O₂ is consumed (Fig. 3, f and h). The kinetics of NF-κB activation by H₂O₂ was much faster with the bolus addition than with the steady-state exposure, reaching a maximum at 60 min, as observed by others in other cell lines (45, 48). This fast activation process is still slower than that achieved by TNF-α and, probably, not physiologically relevant. In fact, 1 mM H₂O₂ is an extremely high concentration that most likely is not even reached in the phagocytic vacuole of polymorphonuclear neutrophils (1). Although no cytotoxicity was observed for TNF-α alone or with [H₂O₂]_ss, there was a significant loss of cell viability (∼35% for MCF-7 cells and 40% for HeLa cells) for 1 mM H₂O₂ bolus addition after 1 h (data not shown).

We hypothesized that the high oxidation levels imposed by the H₂O₂ bolus addition to cells could also be mimicked by a longer incubation with a lower [H₂O₂]_ss. To test this hypothesis MCF-7 cells were preexposed to a [H₂O₂]_ss of 25 μM for 3 h before the addition of TNF-α (Fig. 3,i). Under these conditions, H₂O₂ had an antagonistic effect on TNF-α-induced p65 translocation, with an inhibition of up to 60% (Fig. 3 j), but contrary to the bolus addition, there was a sustained antagonism and cell viability was maintained (not shown).

Overall, our results indicate that H₂O₂ has a modulatory effect on NF-κB activation by TNF-α. Under conditions of inflammation where TNF-α and H₂O₂ are simultaneously present, a positive modulatory role for H₂O₂ in TNF-α-induced NF-κB activation is predicted. Nevertheless, this modulatory role is dependent on the cellular redox status and may switch from stimulatory to inhibitory thus explaining the contradictory results found in the literature.

To address the importance of the positive modulatory role of H₂O₂ on the activation of NF-κB by TNF-α, we analyzed with microarrays the expression of the NF-κB-dependent genes listed in Table IA of supplemental material.⁴ The action of H₂O₂ was studied using both a short (1 h) and a long (6 h) time of exposure to mimic the beginning of an inflammatory situation, when residential macrophages in the tissue are activated and proinflammatory cytokines are liberated—immediate innate immunity—and the progression to an acute inflammatory response, which includes the additional recruitment of effector cells and induction of acute phase proteins (49).

After 1 h of exposure to TNF-α, many genes were induced in MCF-7 cells (Table IB of supplemental material) namely its own mRNA (TNF-α); the chemokines MCP-1 and IL-8 that attract leukocytes to site of inflammation; and genes of the NF-κB pathway, such as IκB-α and the zinc-finger protein A20, both responsible for the negative feedback of NF-κB activation. HeLa cells were also very responsive, with high expression of IL-8, TNF-α, A20, and IκB-α, as observed for MCF-7 cells. In addition, HeLa cells expressed the cytokine IL-6, which has an important role for resolution of the inflammation process and initiation of the immune response (50). Thus, the MCF-7 and HeLa epithelial cell lines have a functional NF-κB pathway, and are able to switch on the production of cytokines and chemokines, as it would be expected at the beginning of an inflammatory response for an immunological active cell. This has been previously observed for other epithelial cell lines and is in agreement with an active role for epithelial cells during the innate immune response (51).

After 1 h, a [H₂O₂]_ss of 25 μM did not induce genes in HeLa cells and induced only a limited number of genes in MCF-7 cells, with more relevance for the cytokine TNF-α, the adhesion molecule ICAM-1 and the transcription factor Jun that belongs to the AP-1 family (Table I, B and C of supplemental material). This is consistent both with the absent or low translocation of p65 into the nucleus induced by H₂O₂ at 1 h (Fig. 3, a and c) and the fact that the induction observed for the three genes could be due to activation of transcription factors other than NF-κB (52, 53, 54). Concerning the modulation of the activation of NF-κB by TNF-α, Fig. 4 shows the relationship between gene expression induced by the simultaneous addition of H₂O₂ plus TNF-α and gene expression induced by only TNF-α. Only two genes in MCF-7 and in HeLa cells were modulated significantly by H₂O₂, and therefore, it can be concluded that H₂O₂ does not have a regulatory role in the modulation of TNF-α action at 1 h (Fig. 4, c and d, and Table II).

After 6 h of incubation in both MCF-7 and HeLa cells, TNF-α still led to an increased expression of genes related with the inflammatory/immune response and the NF-κB pathway (Table I, B and C, of supplemental material), as expected from the oscillatory prolonged p65 translocation caused by the continuous presence of TNF-α (Fig. 2,e). Concerning H₂O₂, and in contrast to what was observed at 1 h of incubation, a [H₂O₂]_ss of 12.5 μM induced several genes in MCF-7 cells (Table IB of supplemental material), which seems to indicate a late effect of H₂O₂ and is consistent with the slow kinetics of p65 nuclear translocation induced by H₂O₂ observed in Fig. 2,d. In particular, heme oxygenase-1 (HO-1), ICAM-1, IFN-α, the putative NF-κB-activating protein HNLF and early growth response-1 (Egr-1) expressions were higher in the presence of H₂O₂ than in the presence of TNF-α at 6 h. In HeLa cells, H₂O₂ also activated some genes after 6 h, especially EGR1, IL-6, ICAM-1, and A20. However, some of these genes are probably being expressed under the control of other transcription factors (for example, HO-1 is probably activated by H₂O₂ via NF-E2-related factor-2 (55)). More importantly, when comparing gene induction caused by TNF-α with gene induction caused by H₂O₂ plus TNF-α, it can be seen that an important subset of genes involved in the regulation of the inflammatory response (HO-1, G-protein-coupled receptor 89 (GPR89), ICAM-1, IL-8, IFN-α, MCP-1, TLR2, and TNF-α in MCF-7 cells and IL-6, IL-8, HO-1, and EGR1 in HeLa cells) is positively modulated by H₂O₂ (Fig. 4, e and f). Remarkably, by itself H₂O₂ did not induce significantly some of these genes (GPR89, IL-8, MCP-1, TLR2), which emphasizes the synergistic action of H₂O₂ on TNF-α activation of NF-κB (Table II).

In conclusion, in conditions mimicking inflammation and in the presence of TNF-α, H₂O₂ has a modest modulatory role after 1 h because only a few genes have an increased expression; however at 6 h, H₂O₂ has an important modulatory role as shown by the increased expression of many genes, mainly those involved in the inflammatory response.

The question of whether H₂O₂ stimulates or inhibits NF-κB has been under dispute because opposite results have been reported. A positive effect may exacerbate inflammation leading eventually to inflammation-related diseases (16, 20) whereas an inhibitory effect is anti-inflammatory and may be viewed as protecting from chronic inflammation or sepsis (21, 56, 57). This depiction may be simplistic because it may be advantageous to stimulate inflammation to have a strong defense response, whereas a putative anti-inflammatory action could be negative by blocking this defensive mechanism. In this work, a switch from a stimulatory to an inhibitory role was achieved by changing the way H₂O₂ was delivered to cells. Although a bolus addition or a pretreatment with [H₂O₂]_ss of 25 μM for 3 h led to a strong inhibitory effect, the simultaneous addition of the same [H₂O₂]_ss with TNF-α caused a synergistic effect. A pretreatment with H₂O₂ has been interpreted by some investigators as representative of chronic inflammation because cells have time to adapt to H₂O₂ (17). Cellular adaptation to H₂O₂, among other alterations, includes increasing the activity of H₂O₂ removal enzymes (40), leading to an increased consumption of H₂O₂ by cells and a consequent decrease in H₂O₂ levels during the experiments. During the incubations with H₂O₂, steady states obtained were stable and did not require further additions of glucose oxidase, indicating that under our experimental conditions H₂O₂ removal enzymes were not inactivated and that cells were not adapted to H₂O₂. In fact, the time required for adaptation to H₂O₂ is around 18 h (58), while “permanent” long-term adaptations would require for treatments with H₂O₂ to last for weeks (59), which contrasts with our up to 6 h shorter exposures. Therefore we think that H₂O₂ role in TNF-α-induced NF-κB activation switches from stimulatory to inhibitory due to an increase in the oxidative load and not due to an adaptative response of the cells. This is consistent with the dual regulation of NF-κB by the redox state of the cell (23, 24, 25).

Acute inflammation starts ∼4 h after infection or trauma, peaks around 24 h lasting up to 4 days and, in an animal model of LPS-induced cystitis, NF-κB-dependent gene induction peaks around 4 h (60). Under our conditions, after a 6 h incubation with a [H₂O₂]_ss of 12.5 μM, a concentration that can be found in vivo under inflammatory conditions after neutrophil activation (14, 15), H₂O₂ clearly stimulated the expression of many NF-κB-dependent genes in MCF-7 cells, induced by TNF-α. Besides some regulatory genes of the NF-κB pathway, most genes stimulated by H₂O₂ have a proinflammatory role namely the adhesion molecule ICAM-1, the proinflammatory cytokine TNF-α, the chemokines MCP-1 and IL-8, and the chemokine receptor GPR89. TLR2 and IFN-α, which are important for antibacterial and antiviral protection, respectively, were also up-regulated by H₂O₂.

It is also highly relevant that the increased expression of these proinflammatory genes was accompanied by the activation of the HO-1 gene, which was observed in both cells lines studied, with a higher expression in MCF-7 cells. HO-1 controls the oxidative degradation of heme, forming as an end product carbon monoxide (CO), which has an anti-inflammatory function (61). In LPS-activated macrophages, CO mediates the inhibition of the production of proinflammatory cytokines, such as IL-1b, TNF-α, and chemokines, such as MIP-2 (61). HO-1 also inhibits ICAM-1 in endothelial cells (62). The induction of HO-1 by H₂O₂ may not involve NF-κB, but rather the NF-E2-related factor-2 (55); nevertheless, the high expression of this gene after 6 h of exposure to H₂O₂ can potentially fulfil a negative feedback role to avoid an excess of proinflammatory conditions. In HeLa cells, this anti-inflammatory role of H₂O₂ may be mediated also through stimulation of IL-6. After the initiation of a local and systemic inflammation IL-6^−/− mice showed a higher expression of proinflammatory cytokines, such as TNF-α and MIP-2, than IL-6^+/+ mice, which indicates an anti-inflammatory role for IL-6 (63). It is also interesting to note that in MCF-7 cells, where this cytokine was not induced, H₂O₂ fulfilled its putative anti-inflammatory role through a higher stimulation of HO-1.

Although the proposal that H₂O₂ has simultaneously a pro- and an anti-inflammatory action may seem contradictory, the biological design in which the activation of a pathway produces an inhibitor which blocks the activation process is a classic feedback mechanism that is very common not only in metabolic pathways, but also in gene networks. For example, this design is found in the pairs NF-κB/(IκB, A20), p53/Mdm2, HSF1/hsp70 and in many others (64). Therefore, the coexistence of two regulatory loops, one proinflammatory in which H₂O₂ induces the production of proinflammatory cytokines, through a synergistic action with TNF-α, and another anti-inflammatory in which H₂O₂ either activates the production of CO through HO-1 or of IL-6, allows H₂O₂ to fine-tune the inflammatory process. Among other anti-inflammatory actions, HO-1 inhibits NF-κB activation (65). Like IκB-α and A20, which act as negative feed-back regulators of NF-κB activation, HO-1 can also act as a negative feed-back regulator of the H₂O₂-dependent synergistic activation of NF-κB. However CO, which is a highly diffusible gas, would not limit its action to the cell of origin having a potential important role in cell-cell communication at the site of inflammation. The anti-inflammatory role for H₂O₂ is supported by the increased IL-8 production found in neutrophils isolated from patients with chronic granulomatous disease (3) and by the deregulated excessive inflammatory responses (such as increased TNF-α and IL-1β levels) observed in animal knockout models of this disease when stimulated with sterile inflammatory stimulus (66, 67). In contrast, the proinflammatory role for H₂O₂ is supported by the decreased formation of MCP-1 observed both in glucan-induced pulmonary granuloma in vivo when H₂O₂ is removed by catalase addition (68), and in a whole animal ischemia model when ROS are removed by antioxidants (69). Furthermore, in vitro, many studies show that H₂O₂ stimulates the formation of chemokines (70, 71, 72, 73, 74).

Due to their localization, epithelial cells are the first to contact the pathogens thus constituting a first defensive physical barrier against them. Besides this passive role, there is recent evidence that epithelial cells participate actively in the inflammatory and innate immune responses, including pathogen recognition mediated by TLR receptors, production of proinflammatory cytokines, chemokines, adhesion molecules and growth factors (51, 75, 76). The results obtained in this work with the two epithelial cell-lines MCF-7 and HeLa further support an active role for the epithelium (Fig. 5). After phagocyte activation, epithelial cells will be subjected to proinflammatory cytokines, such as TNF-α and to exogenous H₂O₂ produced by phagocytes. H₂O₂ will stimulate the activation of NF-κB by cytokines in epithelial cells, producing cytokines and chemokines that will attract more phagocytes to the site of assault, thus amplifying the inflammation. This amplification is homeostatically controlled by the H₂O₂-dependent induction of the HO-1 or IL-6 genes in epithelial cells, which lead to the release of the anti-inflammatory mediators CO and IL-6.

Although this work was conducted with two epithelial cell lines, there are implications for other cell types. The dependency of the regulation of NF-κB activation, either inhibitory or stimulatory, on the level of oxidative stress observed in MCF-7 and HeLa cells, may be a general phenomenon, taking in account the vast literature reporting contradictory findings for NF-κB regulation by H₂O₂. Also the weak capacity of H₂O₂ to activate by itself NF-κB is a result of general importance, because usually large doses of H₂O₂ are needed to observe NF-κB activation. Most probably for most cell types, in vivo H₂O₂ concentrations activate NF-κB weakly, and the main biological role of H₂O₂ is the stimulation of NF-κB activation by other species, such as TNF-α. The individual genes whose expression is increased by H₂O₂ are cell-type dependent, although this work suggests a general pattern in which H₂O₂ activates both pro- and anti-inflammatory genes. We purpose that in addition to its well-established germicide function H₂O₂ has an important dual role as a fine-tuning regulator of the inflammatory process.

The authors have no financial conflict of interest.