Hydrogen peroxide scavenging is not a virulence determinant in the pathogenesis of Haemophilus influenzae type b strain Eagan

Background A potentially lethal flux of hydrogen peroxide (H2O2) is continuously generated during aerobic metabolism. It follows that aerobic organisms have equipped themselves with specific H2O2 dismutases and H2O2 reductases, of which catalase and the alkyl hydroperoxide reductase (AhpR) are the best-studied prokaryotic members. The sequenced Haemophilus influenzae Rd genome reveals one catalase, designated HktE, and no AhpR. However, Haemophilus influenzae type b strain Eagan (Hib), a causative agent of bacterial sepsis and meningitis in young children, disrupted in its hktE gene is not attenuated in virulence, and retains the ability to rapidly scavenge H2O2. This redundancy in H2O2-scavenging is accounted for by peroxidatic activity which specifically uses glutathione as the reducing substrate. Results We show here that inside acatalasaemic H. influenzae all of the residual peroxidatic activity is catalyzed by PGdx, a hybrid peroxiredoxin-glutaredoxin glutathione-dependent peroxidase. In vitro kinetic assays on crude hktE- pgdx- H. influenzae Rd extracts revealed the presence of NAD(P)H:peroxide oxidoreductase activity, which, however, appears to be physiologically insignificant because of its low affinity for H2O2 (Km = 1.1 mM). Hydroperoxidase-deficient hktE- pgdx- H. influenzae Rd showed a slightly affected aerobic growth phenotype in rich broth, while, in chemically defined medium, growth was completely inhibited by aerobic conditions, unless the medium contained an amino acid/vitamin supplement. To study the role of PGdx in virulence and to assess the requirement of H2O2-scavenging during the course of infection, both a pgdx single mutant and a pgdx/hktE double mutant of Hib were assayed for virulence in an infant rat model. The ability of both mutant strains to cause bacteremia was unaffected. Conclusion Catalase (HktE) and a sole peroxidase (PGdx) account for the majority of scavenging of metabolically generated H2O2 in the H. influenzae cytoplasm. Growth experiments with hydroperoxidase-deficient hktE- pgdx- H. influenzae Rd suggest that the cytotoxicity inflicted by the continuous accumulation of H2O2 during aerobic growth brings about bacteriostasis rather than bacterial killing. Finally, H2O2-scavenging is not a determinant of Hib virulence in the infant rat model of infection.


Background
Haemophilus influenzae is a common pathogen among children and immuno-comprised adults with clinical manifestations that are largely type specific. The encapsulated H. influenzae serotype b (Hib) usually causes invasive infections, such as meningitis and septicemia [1], whereas the much more common nonencapsulated, or nontypeable, H. influenzae is a major cause of otitis media, sinusitis, and pneumonia [2]. H. influenzae colonizes the nasopharynx of up to 75% of the population, from where the Hib strains in particular can invade the bloodstream and subsequently pass to the central nervous system. In the course of this pathogenic sequence, the organism moves from sites with high partial oxygen pressure (the nasopharyngeal mucosa; pO 2 = 100 to 160 mm Hg [3]) to lower oxygenated body compartments (arterial and venous blood and cerebrospinal fluid; pO 2 = 3 to 100 mm Hg [3]). These latter levels of oxygen, however, are generally sufficient to inflict injury on colonizing bacteria that are strictly anaerobic in nature or that have been deprived of defences against oxygen toxicity [4][5][6][7].
Molecular oxygen chemically oxidizes redox centers in all aerobic organisms, generating a flux of hydrogen peroxide (H 2 O 2 ) and superoxide radicals (O 2 •-) that can potentially damage the cell through chemical modification of cellular building blocks, in the case of DNA leading to an increased and lethal mutation rate [8]. The array of protective measures oxygen-respiring aerobes developed to deal with H 2 O 2 /O 2 •emphasises the burden that oxidative stress clearly puts on aerobic life. Aerobic organisms generate -or garner from their surroundings -a variety of water-and lipid soluble anti-oxidant compounds. Additionally, virtually all oxygen-respiring organisms contain enzymes that convert O 2 •and H 2 O 2 to innocuous products. Moreover, several damage removal/repair enzymes are constitutively synthesized to deal with chemically modified proteins, lipids and DNA. Finally, since O 2 •-/ H 2 O 2 -levels may vary from time to time -because these levels are the result of a first order chemical reaction with respect to oxygen tension [8], and because such stress can also result from an exogenous source (e.g. bacterial competitors [9] or host inflammatory cells) -organisms are able to adapt to such fluctuating oxidative stresses by inducing the synthesis of antioxidant and damage/repair enzymes. Because of the ubiquity of O 2 •-/H 2 O 2 -scavenging enzymes among oxygen-respiring organisms, it follows that scavenging should hold a prominent place among the protective measures against oxygen toxicity. This prediction was affirmed by studies of superoxide dismutase (SOD)-deficient and hydroperoxidase-deficient mutants of Escherichia coli. sodA -sodB - [10] and katG -katE -ahpR - [11]E. coli both suffer elevated rates of oxygen-mediated DNA damage when grown aerobically in rich broth.
By characterizing and comparing the H 2 O 2 -sensitivities of acatalasaemic (katG -katE -) E. coli and E. coli defective in alkyl hydroperoxide reductase (AhpR), Seaver and Imlay [11] found that the H 2 O 2 dismutase activity of catalase and the H 2 O 2 peroxidatic activity of AhpR serve redundant, but distinct roles inside the E. coli cell. The bipartite alkyl hydroperoxidase system AhpR, composed of a typical 2-cys peroxiredoxin AhpC (the actual peroxidase) and the flavoprotein reductant AhpF [12], are the primary scavengers of endogenous low level H 2 O 2 , while catalase is the more effective scavenger when H 2 O 2 -levels are high and, presumably, when the absence of a carbon source depletes the cell of NAD(P)H necessary for AhpR activity.
Compared to competing bacteria in the upper respiratory tract of humans, H. influenzae is generally more sensitive to either oxygen-and H 2 O 2 -mediated cytotoxicity [4,9], which may corroborate with the predicted absence of an AhpR homologue in the sequenced H. influenzae Rd genome [13]. Nonetheless, mutating the sole structural gene for catalase, designated hktE, does not cause H. influenzae type b strain Eagan to grow poorly under aerobic conditions, nor to be reduced in virulence [14]. In fact, acatalasaemic H. influenzae Rd compared to its parental strain is not significantly more sensitive to the antimicrobial H 2 O 2 -production of Streptococcus pneumoniae [9], implying that catalase does little to protect H. influenzae under these conditions. Additional and efficient hydroperoxidase activity was thus envisioned to be expressed by H. influenzae, and recently a candidate structural gene, termed pgdx, was cloned and characterized [15]. pgdx encodes for the atypical 2-cys peroxiredoxin PGdx, which catalyzes the reduction of both H 2 O 2 and organic hydroperoxides, specifically by using glutathione -which H. influenzae has to garner from its surroundings [16] -as the reducing substrate. Besides its probable role as central H 2 O 2 -scavenger, Murphy et al. [17] proposed a function for PGdx in the process of biofilm formation of non-typeable H. influenzae during respiratory tract infections, and showed that chronic obstructive pulmonary disease patients persistently colonized with H. influenzae can develop antibodies against PGdx.
In this study, a hktE/pgdx H. influenzae Rd mutant unable to produce either catalase or PGdx was constructed and evaluated with respect to its sensitivity towards endogenously generated and exogenously supplied H 2 O 2 . These mutations were moved to Hib strain Eagan, the prototypic virulent strain used to assess virulence utilizing the infant rat model, to explore the impact of increased H 2 O 2 -stress on the ability of H. influenzae to cause invasive disease. Virulence of the hktE/pgdx double mutant was compared with that of the isogenic pgdx single mutant strain, and with wild-type virulence established for the Hib strain Eagan parent.

Construction of a hydroperoxidase-deficient H. influenzae Rd mutant
On the basis of earlier reports [15,18,19], we suspected that the remainder of hydroperoxidase activity in the acatalasaemic H. influenzae Rd mutant AB2593 (Rd hktE::mini-Tn10Cm) could be attributed to the atypical 2-Cys peroxiredoxin PGdx. To explore this hypothesis further, we constructed a derivative of AB2593 in which the pgdx gene is insertionally inactivated by an ampicillin resistance cassette (see Methods). One chloramphenicol/ ampiccillin resistant colony was isolated and termed hktEpgdx -H. influenzae Rd from here on. Allelic exchange as a result of a double cross over event and the consequent lack of PGdx expression was confirmed by PCR and Western blotting respectively (Fig. 1).
Because of the probable H 2 O 2 -stress provoked by the disruption of both the hktE and pgdx genes, pgdx disruptants of AB2593 were selected under anaerobic conditions. Unexpectedly, however, hktEpgdx -H. influenzae Rd colonies were indistinguishable from wild-type colonies when grown aerobically on sBHI plates. On MIc plates, aerobi-cally grown hktEpgdx -H. influenzae Rd colonies were on the contrary roughly edged and much smaller than their wild-type counterparts (data not shown).
In order to assess residual hydroperoxidase activity inside hktEpgdx -H. influenzae Rd cells, the rate of H 2 O 2 -dissipation was measured in a reaction mixture containing 1.5 µM of H 2 O 2 and 3 × 10 8 whole cells of either the wild-type Rd or the hktEpgdx -H. influenzae Rd strain (Fig. 2). While wild-type cells removed H 2 O 2 to levels beneath the limit of detection within twenty minutes, no H 2 O 2 -turnover was noticeable in the reaction mixture containing hktEpgdx -H. influenzae Rd cells, indicating that the double mutant is unable to remove low micromolar concentrations of H 2 O 2 from solution.

Low micromolar H 2 O 2 -toxicity is bacteriostatic rather than bacteriocidal to hktEpgdx -H. influenzae Rd
Aerobically grown hydroperoxidase-deficient katG -katE -ahpR -E. coli does not survive repeated subculturing in rich Luria-Bertani broth [11]. More precisely, growth rates and final densities are reduced after each dilution, which indicates H 2 O 2 -stress mediated DNA damage [11]. Because of our observation that the hktEpgdx -H. influenzae Rd double mutant is indistinguishable from its isogenic parent when cultured aerobically on sBHI solid media, we won-Western blotting and PCR analysis are consistent with the integration of an ampicillin resistance cassette into the chromo-somal pgdx gene of the acatalasaemic H. influenzae strain AB2593 and the virulent hktE -H. influenzae type b strain Eagan mutant   3A shows the aerobic growth curve of wild-type and hktEpgdx -H. influenzae Rd cells in sBHI broth. The aerobic growth defect of the hydroperoxidase-negative strain is marginal, though significant, characterized by a slightly lowered doubling time and a reduction in final cell density. More precisely, upon the entry of the early stationary phase of growth, doubling of the hktEpgdx -H. influenzae Rd culture rapidly stopped. However, no significant decrease in the number of viable cells was apparent when comparing dilutional plating of early stationary phase cultures with overnight cultures (data not shown).
In a previous report, we showed that, because catalase is less efficient in scavenging metabolically-generated H 2 O 2 than is PGdx, catalase of H. influenzae Rd cells that lack functional PGdx (because of the absence of its reductant, i.e. glutathione) is induced about two-fold during routine aerobic growth compared to catalase activity of totally hydroperoxidase-proficient cells [18], while a transient induction of about seven-fold is noticeable after an anaerobic culture was shifted to air [19]. This observation favours the general view that the oxidative stress encountered by microorganisms during an aerobic shift is substantially higher than during routine aerobic growth. So, wild-type and hktEpgdx -H. influenzae Rd cells were grown anaerobically in sBHI broth to early exponential phase, after which the cultures were shifted to air (Fig. 3B). Hydroperoxidase-deficiency once more did not bring about a fundamental reduction in growth rate in response to the applied oxidative stress, while again lowered the final culture density compared to the isogenic parent. From these anaerobic-to-aerobic shifted cultures, dilutions were prepared at the early stationary phase, and aerobic growth was monitored (Fig. 3B). The resulting growth curves are similar to those obtained from anaerobically pregrown cultures, indicating that the higher level of oxidative stress encountered during the shift to aerobiosis does not cause significant DNA damage to hydroperoxidase-deficient H. influenzae Rd grown in rich medium.
On the other hand, growth experiments in chemically defined MIc medium show that under these conditions strain hktEpgdx -H. influenzae Rd is highly vulnerable to oxidative stress. No growth was observed in fully aerated cultures after dilution ( Fig. 4A) or after shifting an anaerobic early exponential phase culture to air (Fig. 4B). Moreover, severe growth retardation was noticeable under the microaerophilic conditions present in a non-shaking candle extinction jar (data not shown). The underlying cause of the encountered oxidative stress is likely to be the accumulation of H 2 O 2 , since the addition of puryvate (a nonenzyme scavenger of H 2 O 2 [20]) or catalase to aerobically growing hktEpgdx -H. influenzae Rd cultures resulted in wild-type growth (Fig. 4B).
In 1976, Boehme and coworkers [21] reported that certain amino acid biosynthesis pathways of E. coli are extremely vulnerable to oxidative inactivation. To test the hypothesis that aerobically grown hktEpgdx -H. influenzae Rd cells in MIc medium are not viable because of the inability to synthesize certain amino acids or vitamins, the growth experiments were repeated in MIc medium that was supplemented with all 20 essential amino acids (final concentration of 40 µg/ml), together with the vitamins riboflavin, niacinamide, pyridoxine and thiamine (final concentration of 1 µg/ml) (Fig. 4A). The aerobic growth defect of the hktEpgdx -H. influenzae Rd strain was largely alleviated by the amino acid/vitamin enrichment of the minimal medium, as inferred from the slightly lower doubling time compared to growth in rich sBHI broth.
The growth experiments described above are based on the cytotoxicity of low micromolar concentrations of H 2 O 2 which are inevitably generated during oxygen-respiration [8]. The effect of higher levels of H 2 O 2 -which e.g. could be the result of the antimicrobial repertoire of bacterial competitors [9] or host phagocytes -on the fitness of the hydroperoxidase-deficient hktE pgdx double mutant, was assessed via disk diffusion testing (Fig. 5 influenzae Rd, the wild-type strain was also heavily sensitized to the cytotoxicity of the supranormal levels of H 2 O 2 when assayed on MIc agar plates. In summary, respiratory-generated H 2 O 2 seems to affect growth of hydroperoxidase-deficient H. influenzae Rd by blocking the supply of cellular building blocks, resulting in bacteriostasis. It thus appears that hktEpgdx -H. influenzae Rd prevents the continuously generated stream of H 2 O 2 being bacteriocidal, either by limiting the ferrous iron-mediated chemical reduction to the extremely harmful hydroxyl radicals (Fenton chemistry [22,23]) or perhaps by having particularly efficient DNA damage repair mechanisms.

Crude hktEpgdx -H. influenzae Rd cell extracts contain low-specific NAD(P)H peroxidase activity
Because PGdx reduces peroxides, while concomitantly oxidizing glutathione, direct spectrophotometric monitoring of in vitro PGdx activity is feasible by following the NADPH-dependent reduction of glutathione disulfide catalyzed by the flavoprotein glutathione reductase. By using t-butyl hydroperoxide (t-BOOH) as the peroxide substrate, the usefulness of this assay to enzymatically confirm the hktE pgdx double mutation was limited because of severe background activity, i.e. oxidation of the nicotinamide nucleotide reductant was already apparent in reaction mixtures solely containing NADPH, t-BOOH and crude extract. Similar NADPH-dependent peroxidase activity was observed for mixtures containing NADPH, H 2 O 2 and crude extract. Because this in vitro peroxidatic activity conflicts with the previous conclusion that hktEpgdx -H. influenzae Rd cells are totally devoid of H 2 O 2 -scavenging activity, we determined the kinetic parameters of the NADPH peroxidatic activity using crude extracts (Fig.  6). The saturation curves for the reducing substrates NADPH and NADH (Fig. 6A) yielded comparable specificities, with K m values resembling their in vivo concentrations (K m -NADPH = 37.4 µM; K m -NADH = 55.0 µM). On the other hand, the K m values for the oxidizing substrates t-BOOH (K m -t-BOOH = 5.6 mM) and H 2 O 2 (K m -H 2 O 2 = 1.1 mM) are far above physiologically relevant in vivo concentrations (Fig. 6B). In fact, the affinity constant for H 2 O 2 is 3 orders of magnitude higher compared to the K m value of ~2 µM for PGdx catalyzed H 2 O 2 -reduction [15], likely explaining that this novel NAD(P)H:peroxide oxidoreductase activity is of minor importance when assaying the turnover of low micromolar concentration of H 2 O 2 by whole hktEpgdx -H. influenzae Rd cells.
For the NADPH:peroxide oxidoreductase activity to be a determinant factor for the hktEpgdx -H. influenzae Rd cells remaining as fit under aerobic conditions in rich medium as their isogenic parent, one would expect this peroxidatic activity to be regulated in response to oxidative stress. So, The scavenging of endogenously generated H 2 O 2 is not a pre-requisite for aerobic growth of H. influenzae Rd in rich com-plex medium NADPH peroxidatic activity was measured in crude extracts derived from wild-type and hktEpgdx -H. influenzae Rd cultures using t-BOOH as the oxidizing substrate. For the purpose of the present study, monitoring the NADPH-dependent turnover of H 2 O 2 would be more appropriate, however, the endogenous catalase activity of wild-type cells conflicts with this approach. No induction of NADPH-dependent t-BOOH peroxidase activity was apparent. In fact, from the recorded specific activities of 11.8 ± 1.3 nmol/min mg protein and 8.2 ± 2.1 nmol/min mg protein for wild-type and hktEpgdx -H. influenzae Rd cell extracts respectively, it seems that NADPH:t-BOOH oxidoreductase activity is slightly repressed as the result of H 2 O 2 -stress.

Hydroperoxidase-deficient H. influenzae strain Eagan is not attenuated in virulence
To determine the involvement of PGdx in the pathogenic sequence leading to bacteremia and to assess the influence of increased H 2 O 2 -stress on Hib virulence, both a single Hib Eagan mutant disrupted in its pgdx gene, and a Hib Eagan hktE pgdx double mutant were created by moving these mutations from the genetically modified Rd strains to competence-induced Hib Eagan cells (Fig. 1). Growth of wild-type and hktEpgdx -Hib Eagan strains in rich sBHI and chemically defined MIc liquid medium showed similar trends as described for their Rd counterparts (data not shown). Moreover, Fig. 2 shows that the hktEpgdx -Hib Eagan double mutant is completely unable to metabolise low µM concentrations of H 2 O 2 , while Fig. 5 shows that the double mutant cells are sensitised to H 2 O 2 -stress due to low-complex nutrient availability. These two phenotypes match those observed for the Rd counterparts and conclusions are to be drawn accordingly. To assess virulence, wild-type Hib Eagan and mutant strains were cultured anaerobically to mid-exponential phase, diluted tõ 200 CFU/100 µl, and intraperitoneally inoculated into 5-day-old infant rats. Bacteremia was assessed at 48 hours by culturing a tail vein blood sample anaerobically on sBHI plates containing the appropriate antibiotics. Compared to wild-type Hib Eagan (814 ± 380 CFU/5 µl blood; n = 6), both pgdx -(899 ± 248 CFU/5 µl blood; n = 5) and hktEpgdx -Hib Eagan (782 ± 433 CFU/5 µl blood; n = 4) were not attenuated in virulence.

Discussion
A Hib mutant defective in lipoamide dehydrogenase is indistinguishable under anaerobic conditions from its isogenic parent, while showing no growth at all in the presence of air [24]. Because this strain is severely reduced in virulence, it was concluded that Hib requires the ability for aerobic respiration in order to complete its pathogenic sequence leading to invasive disease. It thus follows that Hib strains are subjected to O 2 •-/H 2 O 2 -toxicity in the course of bacteremia. Nonetheless, the fitness of our hydroperoxidase-deficient H. influenzae Rd mutant is only marginally affected by aerobic conditions in rich sBHI broth, and hktEpgdx -Hib Eagan displays a normal ability to produce persistent bacteremia in infant rats. Similar observations are reported for SOD-deficient Hib [4], suggesting that in the infant rat model of infection, neither phagocytic cells and their respiratory bursts, nor the potentially toxic O 2 -levels in the blood, play a major role in limiting Hib virulence. The pathogenic sequence of Sal- The scavenging of endogenously generated H 2 O 2 is a prereq-uisite for aerobic growth of H. influenzae Rd in chemically defined minimal medium monella typhimurium -another causative agent of bacteremia in humans -to cause infection when injected into mice by the intraperitoneal route, is also indifferent to the presence of either catalase [25], AhpC or OxyR [26] (which regulates transcription of about 30 proteins in response to fluctuating H 2 O 2 -levels [27]), while a severe attenuation in virulence is noticeable for a Salmonella typhimurium recA mutant defective in DNA repair [25]. Thus, the ability to repair damaged DNA appears to be more important than the ability to directly inactivate the mediators of oxygen toxicity, O 2 •and H 2 O 2 , during the course of invasive infection of pathogenic agents that are able to deceive the host inflammatory system. Acatalasaemic H. influenzae Rd does not grow in chemically defined MIc medium, because of an abrogated ability to remove H 2 O 2 [19]. Because wild-type growth as well as wild-type H 2 O 2 -scavenging activity (of low micromolar levels of H 2 O 2 ) is regained simply by adding glutathione to the minimal medium, we envisioned that the remainder of hydroperoxidase activity inside acatalasaemic H. influenzae Rd is catalyzed in a glutathione-dependent manner [19]. We hypothesized PGdx to be the most likely candidate to catalyze this activity because of its high specificity for H 2 O 2 -reduction (k cat /K m = 5.01 × 10 6 s -1 M -1 ) and because completion of its peroxidatic cycle exclu-sively depends on the presence of glutathione (e.g. thioredoxin can not act as a PGdx reductant) [15]. This hypothesis is confirmed here given that either a H. influenzae Rd or a Hib Eagan double mutant that lacks both catalase and PGdx can not catalyze turnover of micromolar amounts of H 2 O 2 (Fig. 2). Thus, catalase (HktE) and a sole peroxidase (PGdx) account for the majority of H 2 O 2scavenging in the H. influenzae cytoplasm. The H 2 O 2 -scavenging machinery of E. coli also is embodied by catalase (KatG) and a sole peroxidase, in this case AhpC instead of PGdx [11]. Although being clearly dissimilar with regard to the reductive branch of their peroxidatic cycles, the peroxiredoxins AhpC and PGdx can be regarded as being functionally analogous, since i) their kinetic parameters for either the reduction of H 2 O 2 or t-BOOH are very similar [15,28]; ii) they both appear to be of most importance during routine exponential growth (when the H 2 O 2 -concentrations are low and the supply of reductant is high) [11,19]; iii) they both are among the most abundantly expressed proteins [29]; iv) neither PGdx, in case of Hib as reported here, nor AhpC, in case of S. typhimurium [26] as well as in case of the acatalasaemic anaerobe Porphyromonas gingivalis W83 [30], are required for virulence; and v) they both elicit an immunogenic response when injected into infected models [17,26,31]. The latter observation also shows that virulence and immunity are not necessarily connected, as has also been reported e.g. for the major secretory protein of Legionella pneumophila, which results in strong protective immunity, but is apparently nonessential for virulence [32].
On the basis of the established physiologically relevant affinities for its reducing substrates, NADPH (K m = 37.4 µM) and NADH (K m = 55.0 µM), the NAD(P)H:peroxide oxidoreductase activity, detected here in crude extracts of both hktEpgdx -H. influenzae Rd and its wild-type parent, could be of some relevance for the parasite to control its peroxide levels. The affinities, however, for either the simplest peroxide H 2 O 2 (K m = 1.1 mM) or the organic peroxide t-BOOH (K m = 5.6 mM) are so low as to question whether these peroxides are the real in vivo substrates for the NAD(P)H:peroxide oxidoreductase activity. Moreover, we have shown here in Fig. 2 that hktEpgdx -H. influenzae Rd as well as hktEpgdx -Hib Eagan are totally deprived of H 2 O 2 -scavenging activity, meaning that, not only the so-called NAD(P)H peroxidase, but also other hydroperoxidases potentially expressed by H. influenzae, such as the thiol peroxidases Bcp and Tpx [33], are not involved in scavenging endogenously-generated H 2 O 2 . In this respect, the determination of the kinetic parameters for Bcp and Tpx would be of great value to clarify this issue.
A decade ago, Coves et al. [34] reported the NAD(P)H: H 2 O 2 oxidoreductase activity in cell-free E. coli extracts; The hktE/pgdx double mutants of H. influenzae Rd and Hib Eagan are extremely sensitive to supranormal concentrations of exogenously supplied H 2 O 2 , especially under conditions of low-complex nutrient availability biochemical evidence was provided that such activity was distinct from H 2 O 2 -removal catalyzed by KatG and AhpR. The apparent K m values of the E. coli NAD(P)H peroxidase for its reducing substrates were within 30-40 µM [34], which agree well with our measurements for the H. influenzae Rd NAD(P)H peroxidatic activity. Unfortunately, because contaminating catalase activity biased the specific decomposition of H 2 O 2 , no estimation of the binding affinity for H 2 O 2 of the E. coli NAD(P)H peroxidase was provided and no further comparison with the H. influenzae activity reported here is currently possible.
Hydroperoxidase-deficient (katG -katE -ahpR -) E. coli progressively grows slower in rich Luria-Bertani broth and can only grow for two generations in M9 minimal medium supplemented with all 20 amino acids [11]. On the contrary, hydroperoxidase-deficiency (hktEpgdx -) inflicted in either an Rd or an Eagan background resulted in wild-type growth, generation after generation, in rich sBHI broth, while in chemically defined MIc medium supplemented with all 20 amino acids, a slight aerobic growth defect is manifested as the postponement of nearly wild-type exponential growth to lower-than-wild-type stationary phase cell densities. These results thus imply that endogenouslygenerated H 2 O 2 is bacteriocidal (mutagenic) to E. coli, while being rather bacteriostatic to H. influenzae. This difference in cytotoxic behaviour of H 2 O 2 can not be attributed to quantitative differences, since we have previously reported that aerobically grown H. influenzae Rd cells produce H 2 O 2 at a similar rate (~12.4 µM/s) as has been established for E. coli [11,19]. Two plausible explanations, however, can be put forth to address this issue. First, since H 2 O 2 by itself is not mutagenic, the rate of formation of mutagenic hydroxyl radicals derived from H 2 O 2 adjacent to the genomic DNA molecule of E. coli may be higher compared to that nearby the H. influenzae genome. Secondly, the H. influenzae DNA mismatch repair system may be more efficient in repairing its oxidatively damaged DNA than is the E. coli counterpart, either on the basis of pure kinetics or because of a difference in sensitivity towards H 2 O 2 -mediated inactivation. In this regard, it is interesting to note that the human DNA mismatch repair system for example is highly sensitive to H 2 O 2 -mediated inactivation, even at noncytotoxic levels of H 2 O 2 [35]. The present characterization of hydroperoxidase-deficient H. influenzae Rd and Hib Eagan suggests that the wild-type strains should be highly robust against H 2 O 2 -stress. How, then, can it be explained that H. influenzae is more vulnerable to oxygen toxicity compared to other inhabitants of the human nasopharynx [4,9]? First of all, since we have assessed here cytotoxicity of only one product (H 2 O 2 ) of the reaction of oxygen with the cell's redox centers, the possibility remains that growth of H. influenzae is directly affected by oxygen toxicity because of intrinsic vulnerability against O 2 •-. D'Mello et al. [4] indeed reported that growth of H. influenzae is highly affected by O 2 •--stress, as evidenced by the absence of growth of a SOD-deficient Hib strain under fully aerated conditions. Secondly, taken into account the highly fastidious nature and the highly condensed genome of H. influenzae, O 2 •--and H 2 O 2mediated protein inactivation could result in a number of auxotrophies, which can not be relieved under certain culture conditions. In this respect, we consider the basis and diversity of auxotrophies imposed upon H. influenzae by either O 2 •--or H 2 O 2 -stress to be important topics for further research.
The NAD(P)H:peroxide oxidoreductase activity found in crude extracts of hktEpgdx -H. influenzae Rd cultures has physiologically relevant binding affinities for its reducing sub-strates NADPH and NADH, while being very low specific with regard to the oxidizing peroxide substrates H 2 O 2 and t-BOOH Figure 6 The NAD(P)H:peroxide oxidoreductase activity found in crude extracts of hktEpgdx -H. influenzae Rd cultures has physiologically relevant binding affinities for its reducing substrates NADPH and NADH, while being verylow specific with regard to the oxidizing peroxide substrates H 2 O 2 and t-BOOH. Michaelis-Menten representations of NAD(P)H:peroxide oxidoreductase activity with respect to varying concentrations of (A) reductants and (B) oxidants. Kinetic constants were calculated by fitting the data to the Michaelis-Menten equation using a non-linear curve fit. The amount of total protein used for each kinetic assay was 2.5 mg.

Conclusion
By generating mutants of H. influenzae Rd and the virulent strain Hib Eagan defective in both HktE and PGdx, we were able to show that these two hydroperoxidases, a catalase and a peroxiredoxin, account for the majority of scavenging of metabolically-generated H 2 O 2 . No other H 2 O 2 -removal activities appear to be of physiological significance. Yet, in vitro kinetic assays revealed that hktEpgdx -H. influenzae Rd still produce NAD(P)H:peroxide oxidoreductase activity. Although this may represent a detoxifying activity, the NAD(P)H:peroxide oxidoreductase appears to be irrelevant for in vivo H 2 O 2 -scavenging because of its high K m -value (1.1 mM) when considering physiological steady-state H 2 O 2 -concentrations (20-150 nM, as estimated for E. coli [36,37]

Materials
Restriction endonucleases were obtained from New England Biolabs (Beverly, MA). DNA purification from gel or solution was carried out using either the Qiaquick DNA Extraction or PCR Purification Kit (Qiagen, Crawley, UK). Ligations were performed using T 4 DNA ligase (Promega, Madison, WI). Plasmid DNA was prepared by the alkaline lysis method on either a small scale [38] or a 30-ml scale using the Qiagen Plasmid Purification Kit.

Media
Brain heart infusion broth (BHI) was prepared from a dehydrate (Difco, Becton Dickinson and Company, Franklin Lakes, NJ) and autoclaved. To this medium, a Haemophilus test medium supplement (Oxoid, Hampshire, UK), containing V-factor (NAD) and X-factor (hemin), was added according to the manufacturers' instructions to prepare sBHI broth. H. influenzae specific minimal medium (MIc medium) was prepared essentially as described by Herriott et al. [39]. The MIc medium used in this study contained 50 µM of oxidized glutathione, unless indicated otherwise, and was supplemented with a Haemophilus test medium supplement according to the manufacturers' instructions. The amino acid/vitamin supplement used in this study was purchased from Athena Environmental Sciences (Baltimore, MD) and is com-posed of 19 amino acids (40 µg/ml; no methionine (note that MIc medium already contains methionine)), the vitamins riboflavin, niacinamide, pyridoxine-HCl and thiamine (10 µg/ml each), magnesium sulfate (240 µg/ml), ferrous sulfate heptahydrate (25 µg/ml) and glucose (4 mg/ml). Oxygen-free media were generated using a Coy chamber (Coy Laboratory products, Inc.). To prepare agar plates, 1.8% agar was added to the sBHI or MIc liquid growth media before autoclaving. Cultures were routinely grown at 37°C in a candle extinction jar without shaking. When appropriate, 2 µg of chloramphenicol and/or 6 µg of ampicillin were added per ml of either liquid or solid H. influenzae culture media. Growth curves were monitored in the absence of antibiotics and starter cultures were always derived from overnight precultures which were diluted 1:50 to 1:100 to an optical density at 600 nm (OD 600 ) of ~0.005. Aerobic growth was monitored under aerobic conditions as described previously [19]. For each growth experiment, three independent experiments were performed with duplicates and the mean of a single representative set of duplicates (± the standard error of the mean (SEM)) is plotted in the figures.

Bacterial strains and growth conditions
For anaerobic-to-aerobic shift experiments, overnight precultures were diluted 1:50 to 1:100 in oxygen-free growth medium to an OD 600 of ~0.005 and these subcultures were then grown anaerobically to an OD of ~0.15. Anaerobic cultures were prepared in a Coy chamber under an atmosphere of 85% N 2 -10% H 2 -5% CO 2 . The cultures were subsequently shaken (200 rpm) in atmospheric conditions and OD 600 readings were recorded at one hour intervals as described previously [19]. In the case of aerobic-shift growth experiments in the presence of nonenzyme or catalase based H 2 O 2 scavenging, pyruvate (added from a buffered sterile stock solution to a final concentration of 0.75%; Sigma-Aldrich, St. Louis, MO) or