NoxE NADH Oxidase and the Electron Transport Chain Are Responsible for the Ability of Lactococcus lactis To Decrease the Redox Potential of Milk

ABSTRACT The redox potential plays a major role in the microbial and sensorial quality of fermented dairy products. The redox potential of milk (around 400 mV) is mainly due to the presence of oxygen and many other oxidizing compounds. Lactococcus lactis has a strong ability to decrease the redox potential of milk to a negative value (−220 mV), but the molecular mechanisms of milk reduction have never been addressed. In this study, we investigated the impact of inactivation of genes encoding NADH oxidases (noxE and ahpF) and components of the electron transport chain (ETC) (menC and noxAB) on the ability of L. lactis to decrease the redox potential of ultrahigh-temperature (UHT) skim milk during growth under aerobic and anaerobic conditions. Our results revealed that elimination of oxygen is required for milk reduction and that NoxE is mainly responsible for the rapid removal of oxygen from milk before the exponential growth phase. The ETC also contributes slightly to oxygen consumption, especially during the stationary growth phase. We also demonstrated that the ETC is responsible for the decrease in the milk redox potential from 300 mV to −220 mV when the oxygen concentration reaches zero or under anaerobic conditions. This suggests that the ETC is responsible for the reduction of oxidizing compounds other than oxygen. Moreover, we found great diversity in the reducing activities of natural L. lactis strains originating from the dairy environment. This diversity allows selection of specific strains that can be used to modulate the redox potential of fermented dairy products to optimize their microbial and sensorial qualities.

Lactococcus lactis is a lactic acid bacterium that is widely used as a starter in the dairy industry. Its major role is to acidify milk, but it is also one of the most reducing lactic acid bacteria and produces a very negative redox potential (E h ) value in milk (5,8). The E h is indicative of the ability of a system to donate or accept electrons and is strongly dependent on levels of dissolved oxygen (39). The E h affects the microbiota of many dairy products. For instance, it affects the development and activity of the secondary flora in cheese (especially nonstarter lactic acid bacteria) (4,33) and the viability of probiotic bacteria, such as Lactobacillus acidophilus or bifidobacteria, in yogurt or milk (3, 10-12, 20, 27). Moreover, the E h affects the production and stability of aroma compounds (16,23,24). Oxygen is also responsible for formation of the brown discoloration that affects some cheeses and results from the oxidation of a tyrosine catabolite produced by certain yeasts (9). Finally, the oxygen in the medium and/or the E h affects the development of spoilage microorganisms (7) and the expression of virulence factors in some pathogens, such as Bacteroides fragilis or Staphylococcus aureus (6,14,37). The E h and oxygen thus play a major role in the microbial and sensorial qualities of fermented dairy products.
The ability of L. lactis to decrease the E h of milk is thought to be related mainly to the reduction of O 2 (21). Despite the fact that L. lactis has the genes encoding the electron transport chain (ETC), it does not respire unless heme is added to the growth medium, because it lacks heme biosynthesis genes (15).
In the absence of heme, the reduction of O 2 to H 2 O is likely due to the NADH oxidase (NOX) NoxE (21,26). L. lactis also contains the alkyl hydroperoxide reductase (AhpR) system, which is composed of an H 2 O 2 -forming NADH oxidase (AhpF) and a peroxyredoxine (AhpC) that reduces H 2 O 2 to H 2 O (17). The AhpR system may also participate in reduction of oxygen to H 2 O. However, the E h of oxygen-free milk (saturated with nitrogen) is still positive (21), while the E h of milk fermented with L. lactis is close to Ϫ220 mV (8,21). This suggests that, in addition to oxygen, L. lactis reduces other oxidizing compounds present in milk. We recently showed that the functional part of the electron transport chain (ETC) of L. lactis, which is essentially composed of menaquinones and membrane NADH dehydrogenases (NoxA and NoxB), is responsible for the reduction of tetrazolium violet (TV) to formazan when TV is added to milk (38). NoxA and/or NoxB transfers electrons from intracellular NADH to cell membrane-associated menaquinones, which reduce TV at the cell surface or in the cell membrane. Rezaïki et al. (35) also showed that menaquinones are involved in the reduction of O 2 , Fe 3ϩ , and, with considerable efficiency, Cu 2ϩ . These results suggest that the ETC may be involved in the reduction of milk, which contains metal cations (USDA National Nutrient Database for Standard Reference). Despite the importance of the E h for the quality of dairy products, the molecular mechanisms involved in milk reduc-tion by L. lactis have never been addressed. In the present study, we investigated the impact of inactivation of genes encoding NADH oxidases (noxE and ahpF) and components of the ETC (menC and noxAB) on the reduction of ultrahightemperature (UHT) skim milk by L. lactis. Our results showed that two distinct and complementary mechanisms are involved in milk reduction. The first mechanism is essential and relies on the elimination of dissolved oxygen, principally by the NoxE NADH oxidase. The second mechanism is oxygen independent and is due to the ETC, which probably reduces oxidizing compounds other than oxygen. Moreover, we found great diversity in the reducing activities of natural L. lactis strains, which makes control of the E h in fermented dairy products through selection and use of strains that reduce more or less conceivable.

MATERIALS AND METHODS
Bacterial strains and culture conditions. All L. lactis strains used in this study are L. lactis subsp. cremoris strains. L. lactis mutant strains were derived from laboratory strain TIL46 and are listed in Table 1. The noxAB mutant was constructed previously (38), and the menC mutant was obtained previously by insertion mutagenesis and selection for an inability to reduce tetrazolium violet (38). Other wild-type (WT) L. lactis subsp. cremoris strains were supplied by CSK Food Enrichment (Ede, the Netherlands). Lactococcus strains were grown at 30°C either in M17 broth (Difco, Detroit, MI) supplemented with glucose at a final concentration of 0.5% (wt/vol) (GM17) or in UHT skim milk (Lescure, Surgères, France). When required, an antibiotic (tetracycline, erythromycin, or chloramphenicol) was added at a final concentration of 5 g ml Ϫ1 . Bacterial growth was estimated by measuring the optical density at 480 nm (OD 480 ), after dilution of 0.1 ml of a culture with 0.9 ml EDTA (5 g liter Ϫ1 , pH 12.0) in the case of milk cultures. Escherichia coli strains were grown at 37°C with agitation at 200 rpm in Luria-Bertani broth (LB) (Difco, Detroit, MI). When required, erythromycin (final concentration, 150 g ml Ϫ1 ), kanamycin (final concentration, 50 g ml Ϫ1 ), or chloramphenicol (final concentration, 50 g ml Ϫ1 ) was added.
DNA manipulations. DNA extraction, PCR, plasmid preparation, and sequencing were performed as described by Sambrook and Russell (36). DNA restriction and modification enzymes were purchased from New England Biolabs (Ipswich, MA) and used as recommended by the supplier. The primers used in this study were synthesized by Eurogentec (Seraing, Belgium) and are listed in Table 2. L. lactis electrocompetent cells were prepared and transformed as described previously by Holo and Nes (18). Plasmid extraction was performed as described by O'Sullivan and Klaenhammer (32). PCR amplification was carried out with an Applied Biosystems 2720 DNA thermal cycler (Courtaboeuf, France), using the Taq DNA polymerase (MP Biomedicals, Illkirch, France) or Phusion DNA polymerase (high-fidelity PCR master mixture; Finnzymes, Finland).
Construction of the ahpF null mutant. The ahpF mutant was constructed by deleting 435 bp of the ahpF gene using double-crossover gene replacement. A 1,741-bp DNA fragment that included the ahpF gene (1,527 bp) was PCR amplified with primers ahpF-up and ahpF-down containing EcoRI and SacII restriction sites, respectively ( Table 2). A 435-bp deletion inside the PCR product was obtained by digestion with AvaII and ligation of the two outside fragments (649 and 657 bp). The truncated fragment was then PCR amplified with primers ahpF-up and ahpF-down and subcloned into the pGEM-T Easy vector (Promega, Madison, WI). After digestion of the subcloning vector with EcoRI and SacII, the resulting truncated fragment (1,289 bp) was cloned into pGϩhost9 (28), and the recombinant vector was introduced by electroporation into L. lactis  TIL46. The double crossovers leading to the expected gene replacement were screened and obtained as described by Biswas et al. (2). Correct chromosomal deletion of the ahpF gene was verified using PCR. Construction of the noxE null mutant. The noxE gene, encoding the waterforming NADH oxidase NoxE, was disrupted in L. lactis TIL46 using a single crossover, resulting in the noxE mutant (Table 1). An internal 644-bp DNA fragment of noxE was PCR amplified with the noxF and noxR primers (Table 2) and subcloned into the pGEM-T Easy vector (Promega, Madison, WI). The resulting plasmid, pGEM-TnoxE (Table 1), was produced in E. coli TG1 (Table  1) and digested by SacI and AatII. The digestion product was then cloned into the nonreplicative pORInewlux vector (13), leading to pORInoxE (Table 1), which was also produced in E. coli and transformed into TIL46 electrocompetent cells. Transformants were selected with erythromycin (5 g ml Ϫ1 ). The disruption of noxE by pORInoxE was confirmed by PCR using the external primers pORIEry, luxAinv, Nox2F, and Nox2R ( Table 2).
Complementation of the noxE mutant. Complementation of the noxE mutant was achieved by cloning the noxE gene into the pJIM2246 multicopy plasmid vector. A 1,728-bp fragment encoding noxE with its putative promoter and terminator was amplified by PCR from TIL46 total DNA with primers nox3F and nox3R ( Table 2) using the Phusion high-fidelity DNA polymerase. This fragment was cloned into pGEM-T Easy, and the resulting plasmid was produced in E. coli TG1 and digested with SpeI and SacII. The fragment was then cloned into the SpeI/SacII-linearized vector pJIM2246. The resulting plasmid, pJIM2246::noxE (Table 1), was produced in E. coli TG1 repA ϩ and used to transform the noxE mutant. The resulting strain was designated noxE-pJIM::noxE. A control strain was constructed by introducing the empty vector pJIM2246 into the noxE mutant (noxE-pJIM).
Cell extract preparation. Cells were harvested from L. lactis cultures by centrifugation (4,000 ϫ g, 15 min, 4°C). The cell pellets were washed twice with 5 ml 50 mM potassium phosphate buffer (pH 7.0) and stored at Ϫ20°C until they were used. The cells were resuspended in 50 mM potassium phosphate buffer (pH 7.0) containing 0.6 g of glass beads and disrupted twice for 45 s at 180 V and 4.5 m/s (Fast Prep orbital mixer; MP Biomedicals, Illkirch, France). Each cell extract was recovered by centrifugation (17,400 ϫ g, 20 min, 4°C). The protein concentration was determined using a Coomassie (Bradford) protein assay kit (Pierce, Rockford, IL), and bovine serum albumin (BSA) was used as the standard.
NOX assay. NADH oxidase (NOX) activity was determined by using a 1-ml reaction mixture consisting of 50 mM potassium phosphate buffer (pH 7.0), 0.3 mM EDTA, and 0.2 mM NADH. The reaction was initiated by addition of an appropriate quantity of freshly prepared cell extract. NADH oxidation was monitored by spectrophotometry at 340 nm and 25°C (Uvikon XL; Biotek Instruments, Colmar, France). NOX activity was calculated using the molar extinction coefficient of NADH (6,220 M Ϫ1 cm Ϫ1 ) and was expressed in units per milligram of protein. One unit of NADH oxidase activity is defined as the amount of enzyme which catalyzes the oxidation of 1 mol NADH per min at 25°C. Control experiments without the cell extract were performed to verify the absence of spontaneous NADH oxidation.
Milk fermentation and measurement of pH, dissolved oxygen, and redox potential. All milk fermentations were performed at 30°C in 1-liter fermentors (BioStat Q plus; Sartorius) containing 500 ml UHT skim milk (Lescure, Surgères, France). Milk was inoculated (1%, vol/vol) with a milk culture at an optical density at 480 nm (OD 480 ) of 1.8. The cultures were kept homogeneous by gentle stirring (100 rpm). Aerobic incubation was performed with air in the headspace, which remained in contact with the atmosphere via a sterile filter so that oxygen could diffuse into the medium during fermentation (low-oxygen conditions). Anaerobic cultures were grown in nitrogen-saturated milk with the headspace continuously sparged with nitrogen. During L. lactis growth, acidification and the oxygen concentration were monitored using pH sensors (EASYFERM plus K8; Hamilton) and dissolved oxygen (pO 2 ) sensors (OXYFERM FDA 160; Hamilton). The pH sensors were calibrated using standard solutions with pH values of 4.0 and 7.0. The oxygen sensors were calibrated using water sparged with air (21% oxygen) and nitrogen (0% oxygen). The redox potential was measured using redox electrodes (EASYFERM PLUS K8 Rx; Hamilton) that were cleaned, polished, and verified as described by Jeanson et al. (21). Data were acquired using MFCS/win 2.0 software (B. Braun Biotech International). Acidification data were expressed as the ⌬pH, which is the difference (in pH units) between the initial pH (pH 6.6) and the pH measured at each time. As described by Cachon et al. (8), the measured redox potential (E m ) was converted to E h , with reference to the hydrogen electrode, using the following formula: E h ϭ E m ϩ E r , where E r is 204 mV at 30°C for our Ag/AgCl electrodes. In addition, because redox potential values are pH dependent (19), E h was transformed to the redox potential at pH 7 (E h7 ) with the Leistner-Mirna equation: E h7 ϭ E h Ϫ ␣ ϫ (7 Ϫ pH) (25). The ␣ constant determined for milk experimentally at 30°C was 53. Bacterial growth was estimated each hour by measuring the optical density at 480 nm (OD 480 ) after dilution of 0.1 ml of a milk culture with 0.9 ml EDTA (5 g liter Ϫ1 , pH 12.0). Fermentation experiments with L. lactis TIL46 and the derived mutants were performed at least three times independently, and data were compared using Student's t test. Experiments with CSK strains were performed only once or twice.

RESULTS
Analysis of milk reduction mechanism using knockout mutants. (i) NADH oxidase activity. The NADH oxidase (NOX) activities of TIL46 (WT) and the derived noxE and ahpF mutants grown in GM17 were determined using extracts of cells harvested at the end of the exponential growth phase, when NOX activity was optimal ( Table 3). Disruption of the noxE gene suppressed 95% of the NOX activity. Conversely, the noxE mutant complemented with the pJIM expression vector harboring noxE exhibited an 8-fold increase in activity compared with the WT strain. Finally, inactivation of the ahpF gene did not affect NOX activity.
(ii) Aerobic growth and acidification in milk. The abilities of the different mutant strains to grow in milk and acidify under low-oxygen conditions (gentle stirring at 100 rpm) were examined and compared with the abilities of the WT parent ( Fig. 1A and B). The growth of the noxE mutant did not include an exponential phase, and both the growth rate and the final biomass were significantly (P Ͻ 0.001) reduced compared with the growth rate and the final biomass of the WT strain. The acidification rate was also significantly affected (P Ͻ 0.001) but less than the growth rate, suggesting that the noxE mutant produced more lactic acid than the wild-type strain. The growth and acidification kinetics were restored when a noxE expression vector (pJIM::noxE) was introduced into the noxE mutant (see Fig. S1 in the supplemental material), indicating that the growth defect was actually due to the absence of NOX activity. The growth and acidification rates of the ahpF mutant were also strongly reduced (P Ͻ 0.001) compared with the growth and acidification rates of the WT (Fig. 1), while the menC and noxAB mutants grew and acidified slightly slower than the WT strain but the final biomasses and pH reached about the same levels as those of the WT strain (Fig. 1).
(iii) Dissolved oxygen consumption. The abilities of the WT strain and mutants to consume dissolved oxygen were also examined during milk fermentation under low-oxygen conditions (Fig. 2). The WT strain very rapidly consumed dissolved  Fig. S2 in the supplemental material).
In the ahpF mutant culture the pO 2 level reached zero later than it reached zero in the WT strain culture (6 h for the ahpF mutant versus 4 h for the WT strain), but at the same biomass as the WT strain culture (OD 480 , 0.8). The noxAB and menC mutants consumed dissolved oxygen as fast as the wild-type strain (the pO 2 reached zero at 4 h [OD 480 , 0.8]), but when cultures of these mutants entered the stationary growth phase, the pO 2 level immediately rose to 12% Ϯ 1%, while it rose to 2.5% Ϯ 0.5% for the ahpF mutant and to 1.5% Ϯ 1.5% for the WT strain (Fig. 2). These results indicate that NoxE is principally responsible for O 2 consumption at the beginning of growth and that the ETC (which includes menaquinones and NoxAB) is involved in oxygen consumption during the stationary growth phase.
(iv) Milk reduction under aerobic conditions. The abilities of the WT strain and the mutants to decrease the redox potential (E h7 ) of milk were also investigated during milk fermentation (Fig. 3). The WT strain started to reduce milk slowly from 375 Ϯ 30 mV to 300 Ϯ 25 mV, and as soon as the pO 2 reached zero (at 4 h), the E h7 immediately dropped to Ϫ200 mV and then stabilized at around Ϫ230 Ϯ 10 mV until the end of the exponential growth phase (Fig. 3). The noxE mutant slowly reduced milk, and the E h7 only declined to 260 Ϯ 40 mV, which is in line with the relatively high oxygen level (pO 2 , ϳ10%) in milk. The reducing ability of the noxE mutant was restored by complementation with pJIM::noxE (see Fig. S3 in the supplemental material). The ahpF mutant reduced milk later than the WT, when the pO 2 reached zero (after 6 h), but then the E h7 dropped to the same level that was observed with the WT (Ϫ213 Ϯ 5 mV) (Fig. 3). Finally, the mutants whose ETC was affected (menC and noxAB mutants) reduced milk very gradually to an E h7 value of Ϫ65 Ϯ 50 mV at the end of growth (8 h; OD 480 , ϳ3), although the pO 2 reached zero at the beginning of growth (4 h; OD 480 , 0.8) (Fig. 2). For all strains, the E h7 of milk increased after the end of growth, in line with the increase in the pO 2 . All of these results indicate that NoxE contributes markedly to the reduction of aerated milk via elimination of oxygen. The ETC is also strongly involved in milk reduction, probably by reducing oxidizing compounds other than the oxygen present in milk.
(v) Milk reduction under anaerobic conditions. The abilities of the WT strain and the mutants to grow, acidify, and reduce milk under anaerobiosis were also examined (Fig. 4). The anaerobic growth rate of TIL46 in milk (Fig. 4A) was slightly reduced (P Ͻ 0.05) compared with the aerobic growth rate and was similar to the aerobic growth rate of the noxE mutant (P ϭ 0.17) (Fig. 1), indicating that NAD regeneration via NoxE (under aerobiosis) stimulates the growth of L. lactis TIL46 in milk. Of course, the impact of noxE inactivation on anaerobic growth and acidification was weak because NoxE is inactive in the absence of oxygen. In contrast, the negative impact of ahpF inactivation on anaerobic growth (P Ͻ 0.001) was stronger than the impact on aerobic growth, suggesting that the impaired growth of the ahpF mutant was not due to its NADH oxidase activity. The growth rates (P Ͻ 0.01) and final biomasses (P Ͻ 0.01) of the menC and noxAB mutants were also strongly affected (Fig. 4A). Under anaerobiosis, the E h7 of milk before inoculation was essentially the same as that in the presence of O 2 (350 Ϯ 50 mV). Disruption of noxE or ahpF did not affect the ability of TIL46 to decrease the E h7 of milk (Fig. 4C). The E h7 dropped immediately (before the pH of milk started to decrease) from 350 Ϯ 50 mV to a stable value of Ϫ265 Ϯ 15 mV. With the menC mutant the E h7 dropped 1 h later than it dropped with the WT strain, and it stabilized at a higher value (about Ϫ180 Ϯ 10 mV). Finally, the noxAB mutant reduced milk very gradually, and the E h7 stabilized at the same level that it stabilized with the menC mutant (Ϫ180 Ϯ 10 mV). Thus, our results showed that the NoxA and/or NoxB NADH dehydrogenase and, to a lesser extent, menaquinones are involved in milk reduction by L. lactis TIL46 under anaerobic conditions. Natural diversity of milk-reducing activity in L. lactis. (i) Selection of strains whose ETC activity is affected. A collection of 50 L. lactis strains originating from dairy environments was screened for the ability to reduce tetrazolium violet (TV), which is indicative of ETC activity (38). After inoculation onto MilkAgarTV plates (38), only seven strains were not capable of reducing TV as much as reference strain TIL46. However, the control plate containing a pH indicator showed that six of these strains did not acidify milk either. Only L. lactis strain CSK1019, which grew and acidified milk, reduced TV much less than TIL46, in the same way as the noxAB mutant (not shown).
(ii) Milk reduction by natural L. lactis strains. The abilities of eight natural strains (including seven strains randomly chosen from the CSK collection and strain CSK1019 that reduced TV poorly) to reduce milk under aerobiosis and anaerobiosis were examined.
Under aerobiosis, all of the strains grew in milk and reached stationary phase at between 8 and 14 h; CSK1788 grew the fastest, and CSK1019 grew the slowest (see Fig. S4 in the supplemental material). Cultures of all strains reached an OD 480 close to 3, except for the CSK1019 culture, which reached an OD 480 of ϳ1.8. The oxygen consumption and E h7 evolution varied greatly for the eight strains (Fig. 5). Two of the eight strains (CSK1019 and CSK1382) did not decrease the E h7 of milk, like the noxE mutant (Fig. 5B). As expected, neither of these strains exhibited detectable NADH oxidase activity. All of the other strains reduced the E h7 of milk to Ϫ220 Ϯ 20 mV, but at different rates. In all cases, the E h7 dropped after the pO 2 reached zero (Fig. 5A). Nevertheless, the rate of O 2 consumption was not related to the NADH oxidase activity of the strains. For example, CSK1245 displayed about the same activity as CSK2016 and CSK1378 (0.15 Ϯ 0.04 U mg Ϫ1 ), although it consumed O 2 more slowly (the pO 2 reached zero when the culture OD 480 reached 1.8) than the other two strains (the pO 2 reached zero when the culture OD 480 reached 0.5 to 0.7). In addition, strain CSK2036, which exhibited very low NADH oxidase activity (0.06 Ϯ 0.02 U mg Ϫ1 ), consumed O 2 very rapidly (the pO 2 reached zero when the culture OD 480 reached 0.7). These results suggest that in some L. lactis strains, enzymes other than NoxE are capable of consuming O 2 in the medium. Under anaerobiosis, all of the strains reduced milk to similar extents and like TIL46, except for CSK1019, which reduced milk much more slowly, like the noxAB mutant (Fig. 6). Also, the growth and acidification kinetics of CSK1019 in nitrogen-saturated UHT skim milk were similar to those of the noxAB mutant (not shown).

DISCUSSION
Despite the demonstrated effects of the E h on dairy product quality, starter lactic acid bacteria are still essentially characterized by their acidification activities and not by their abilities to decrease the E h of milk (reduction activities). A few studies have reported that the reducing activity of lactic acid bacteria is species dependent and that L. lactis is one of the most reducing lactic acid bacteria (5,8). Also, Jeanson et al. (21) showed that the E h of milk drops only after all of the dissolved oxygen is consumed by L. lactis. However, the molecular basis for the reducing activity of L. lactis has not been investigated previously. Our study confirmed that elim- ination of O 2 is actually a prerequisite for milk reduction and revealed that another mechanism is necessary for the E h7 to fall to Ϫ220 mV. Indeed, we identified and characterized two major mechanisms involved in milk reduction by L. lactis (Fig. 7). The first mechanism relied on the elimination of dissolved oxygen. We showed that NoxE activity was essential for early O 2 consumption during growth because noxE inactivation hampered a decrease in the pO 2 level to zero and consequently prevented a drop in the E h value. AhpF and the ETC did not appear to contribute notably to O 2 consumption during the early growth phase. Indeed, the observed delay in the disappearance of O 2 with the ahpF mutant was mainly due to the delay in growth, and inactivation of the ETC did not affect O 2 consumption significantly. However, AhpF and the ETC may be responsible for the slight consumption of O 2 observed with the noxE mutant (Fig. 2). The minor role played by AhpF in oxygen elimination was consistent with the low specific activity of the pure enzyme (15 U mg Ϫ1 ) compared to that of NoxE (95 U mg Ϫ1 ) (22) and with the fact that its disruption had no impact on the NADH oxidase activity of cell extracts. During the stationary growth phase, NoxE appeared to be much less active, probably because it is unstable when it is subjected to overoxidation and is inhibited by an acidic pH (22). In this phase, the ETC contributed more markedly to O 2 consumption. Indeed, inactivation of the ETC led to a marked increase in the pO 2 at the end of growth. The contribution of the ETC to O 2 consumption is in line with the ability of menaquinones to reduce O 2 to O 2 ⅐ Ϫ (35) and with the higher levels of menaquinones found in stationaryphase cells than in exponential-phase cells (35). Nevertheless, elimination of O 2 could not explain the drop in the E h7 to Ϫ220 mV because the mutants whose ETC was affected (menC and noxAB mutants) consumed O 2 completely and quite rapidly but did not reduce milk as much as the WT. When the dissolved oxygen level reached zero, E h7 was still positive and then gradually declined to Ϫ65 mV. These results suggest that the ETC, which is made up of menaquinones and NoxA and/or NoxB, reduced oxidizing compounds other than O 2 . These compounds may be metal cations or oxidized sulfur compounds, which can be reduced to free thiols that have a very negative standard potential. Indeed, menaquinones are capable of reducing metal cations and other oxidized compounds, such as TV (35,38). The compounds reduced by menaquinones seemed to be present in milk only under aerobiosis, because under anaerobiosis the menC mutant reduced milk as rapidly as the WT. However, the final E h7 value was higher with the menC mutant than with the WT. Surprisingly, the milk-reducing activity of the noxAB mutant was affected more than the milk-reducing activity of the menC mutant under anaerobiosis, suggesting that NoxA and/or NoxB reduced some compounds directly and without the intervention of menaquinones, probably in the membrane where NoxA and NoxB are localized.
Our conclusion that menaquinones are necessary for reduction of the E h of milk to a very negative value is in agreement with the fact that the most reducing lactic acid bacteria, such as L. lactis and Enterococcus faecalis, produce menaquinones, while Streptococcus thermophilus, which reduces the E h of milk to a much lesser extent (to 8 mV), does not produce these compounds (5,8,31). Moreover, we did not find any menaquinone biosynthesis genes in the sequenced genomes of Lactobacillus helveticus and Lactobacillus delbrueckii subsp. bulgaricus, which reduce milk in the same way as S. thermophilus (5), suggesting that they do not produce menaquinones either. However, several Lactobacillus plantarum strains reduce the E h of milk to a negative value (Ϫ154 mV) (5), while the sequenced strain (WCF1) appears to lack several genes involved in menaquinone biosynthesis. Other mechanisms may be involved in the reducing activity of L. plantarum, or menaquinone production may be strain dependent.
Within L. lactis subsp. cremoris, we observed quite a broad diversity of milk-reducing activities. This diversity appeared to be related in part to NADH oxidase activity because the two strains without detectable NADH oxidase activity did not reduce milk. However, some strains with low NADH oxidase activity eliminated O 2 quite rapidly. This might have been due to the possibility that the NADH oxidase activities of these strains in milk were higher than the activities determined in M17 medium, even though we found that the NADH oxidase activities of TIL46 in cells grown in these two media were similar. In fact, two other genes are predicted to encode NADH oxidases in the genomes of L. lactis subsp. cremoris MG1363 and SK11 (30,40). The first gene, noxC, encodes a 547-residue protein in both genomes, while the second gene, noxD, is a pseudogene in MG1363 (which was derived from TIL46) but encodes a 443-residue protein in SK11. The latter gene may be present in some natural strains and may be expressed differently in M17 and milk. Moreover, certain natural strains may also exhibit other oxidase activities, such as pyruvate oxidase activity, the genes for which are present in both sequenced genomes, or a strong ETC activity, which enable them to consume all of the oxygen present. The diversity of the amounts of menaquinone produced by L. lactis strains in milk cultures has been examined previously (31). The amount produced ranged from 50 to 600 nmol per g of lyophilized culture. This great variation could have an impact on the reducing activities of different strains. However, in the present study, the TV reduction test revealed that the ETC activity of only one strain was affected, probably at the level of the NADH dehydrogenases. This result indicates that all of the strains tested produced enough menaquinone to reduce TV. The NoxE activity of the selected strain CSK1019 was also affected. However, the other natural strain without NoxE activity (CSK1382) and the noxE mutant reduced TV as well as reference strain TIL46, indicating that the TV test does not allow selection of strains on the basis of their oxygen consumption. To conclude, great variation in reducing activity was found in natural L. lactis strains originating from the dairy environment. The broad diversity allows selection of specific strains that can be used to modulate the redox potential of fermented dairy products.