Formate producing capacity provided by reducing ability of Streptococcus thermophilus NADH oxidase determines yogurt acidification rate

Yogurt is made by fermenting milk with 2 lactic acid bacteria, Lactobacillus delbrueckii ssp. bulgaricus ( L. bulgaricus ) and Streptococcus thermophilus ( S. thermophilus ). To comprehensively understand the proto-cooperation mechanism between S. thermophilus and L. bulgaricus in yogurt fermentation, we examined 24 combinations of cocultures comprising 7 fast-or slow-acidifying S. thermophilus strains with 6 fast-or slow-acidifying L. bulgaricus strains. Furthermore, 3 NADH oxidase deficient mutants (Δ nox ) and one pyruvate formate-lyase deficient mutant (Δ pflB ) of S. thermophilus were used to evaluate the factor that determines the acidification rate of S. thermophilus . The results revealed that the acidification rate of S. thermophilus monoculture determined the yogurt fermentation rates, despite the co-existence of L. bulgaricus, whose acidification rate was either fast or slow. Significant correlation was found between the acidification rate of S. thermophilus monoculture and the amount of formate production. Result using ΔpflB showed that the formate was indispensable for the acidification of S. thermophilus . Moreover, results of the Δnox experiments revealed that formate production required Nox activity, which not only regulated dissolved oxygen (DO), but also the redox potential. NADH oxidase provided the large decrease in redox potential required by pyruvate formate lyase to produce formate. A highly significant correlation was found between formate accumulation and NADH oxidase activity in S. thermophilus. In conclusion, the formate production ability provided by the action of NADH oxidase activity determines the acidification rate of S. thermophilus, and consequently regulates yogurt coculture fermentation.


INTRODUCTION
Yogurt is one of the most popular fermented dairy products worldwide, dating back to 6000 B.C. (Weerathilake et al., 2014). Yogurt is made by fermenting milk with 2 lactic acid bacteria, Lactobacillus delbrueckii ssp. bulgaricus (L. bulgaricus) and Streptococcus thermophilus (S. thermophilus). (Food and Drug Administration (FDA), 2017) There is a protocooperation between S. thermophilus and L. bulgaricus (Tamime and Robinson, 2007), which results in a higher acidification rate, a lower pH at the end of fermentation, and the production of exopolysaccharides and aromatic compounds when compared with that of monocultures (de Bok et al., 2011;Herve-Jimenez et al., 2009;Sieuwerts, 2016). Fast acidification in yogurt fermentation is very important for stable production, cost reduction, and the prevention of contamination from environmental microorganisms (Stanojević-Nikolić et al., 2016, Sanlibaba and Gucer, 2015, Shelef, 1994. Clarifying the actual compounds and the state of protocooperation that results in rapid yogurt fermentation is crucial. Various protocooperation factors were reviewed by Tamime et.al (2007) and particularly formate gained more attention as a provided substance from S. thermophilus to L. bulgaricus (Suzuki et al., 1986;Zourari et al., 1992, Trimigno et al., 2020. Similarly, formate that stimulates the growth of S. thermophilus and/or L. bulgaricus was reported with the application of mixed-culture transcriptome profiling by Sieuwerts et.al (2010) as a post genomic research. Additionally, folate (Rao et al., 1984.,Crittenden et al., 2003Sybesma et al., 2003), CO 2 (Galesloot et al., 1968;Driessen et al., 1982;Yamauchi et al., 2019), and fumaric acid (Yamamoto et al., 2021) produced by S. thermophilus are provided to L. bulgaricus; on the other hand, amino acid and peptides produced by L. bulgaricus (its protease and peptidase) are supplied to S. thermophilus (Rajagopal & Sandine, 1990;Thomas & Pritchard, 1987;Zourari et al., 1992).
We have previously reported that yogurt fermentation of S. thermophilus ST1131 and L. bulgaricus LB2038 began after the dissolved oxygen (DO) concentration in milk medium decreased to almost 0 ppm (Horiuchi, 2009) and that the suppression of acid production in coculture by DO in milk was caused by the suppression of formate production in S. thermophilus (Horiuchi and Sasaki, 2011). The results using an H 2 Oforming NADH oxidase (Nox)-defective mutant of S. thermophilus ST1131 revealed that Nox is the major oxygen-consuming enzyme of the bacterium (Sasaki et al., 2014). Yogurt fermentation with the S. thermophilus Δnox mutant and L. bulgaricus LB2038 was significantly slower than with S. thermophilus ST1131 and L. bulgaricus LB2038, and the DO concentrations of the mixed culture did not decrease to less than 2 mg/kg within 3 h. These findings suggest that Nox of S. thermophilus ST1131 contributes greatly to yogurt fermentation, presumably by removing DO from milk.
Formate is a well-known co-factor in yogurt fermentation which accelerates the growth of L. bulgaricus in the absence of pyruvate formate-lyase (PFL) (Derzelle, 2005). PFL is an oxygen-sensitive enzyme that catalyzes the production of formate and acetyl-CoA from pyruvate and CoA (EC 2.3.1.54) (Sawers and Watson, 1998), and PFL requires activation by pyruvate formate-lyase activating enzyme (Crain et al., 2014). PFL of Escherichia coli and Streptococcus mutants are inactivated by oxygen (Knappe et al., 1974;Yamada et al., 1985), but the relation between formate production by PFL and Nox in S. thermophilus is not completely understood. Formate is also required for monoculture of S. thermophilus in milk and is used as a substrate not only for formyltetrahydrofolate synthetase in purine biosynthesis (Derzelle et al., 2005), but also for anaerobic ribonucleotide reductase (RNR) (class III NrdD), which uses formate as the reductant (Levitz et al., 2022).
Previously, we revealed that urease plays a crucial role in yogurt acidification and its deficiency inhibits the fermentative acceleration of protocooperation by examining 5 combinations of 3 fast-or slow-acidifying strains of S. thermophilus with 2 CO 2 -responsive or CO 2 -unresponsive strains of L. bulgaricus (Yamauchi et al., 2019). However, there was a large difference in the fermentation time regardless of the difference in urease activity, thereby indicating that urease activity was not the main factor determining the fermentation time.
As stated previously, there are numerous reports of protocooperation factors associated with metabolites and oxygen related to yogurt acidification; however, these effects on protocooperation vary depending on the specific strains and their different combinations. In this study, we evaluated 24 combinations of cocultures of 7 fast-or slow-acidifying S. thermophilus strains with 6 fast-or slow-acidifying L. bulgaricus strains. Furthermore, 3 nox-deficient mutants (Δnox) and one pflB-deficient mutant (ΔpflB) of S. thermophilus were used to evaluate the factor that determines the acidification rate of S. thermophilus. The aim of this study is to provide comprehensive information regarding the protocooperation between S. thermophilus and L. bulgaricus.

Bacterial strains and growth conditions
The strains used in this study are listed in Table 1. Strains other than ATCC 19258 T and ATCC 11842 T were obtained from stock culture strains of the Food Research and Development Center, Meiji Corporation (Tokyo, Japan).
Each L. bulgaricus and S. thermophilus strain stored at −80°C was pre-cultured at 37°C for 16 h in skim milk and yeast extract (SMY) medium, which consisted of 10% (wt/wt) skim milk and 0.05% (wt/wt) yeast extract (first preculture). The SMY medium was sterilized for use by autoclaving (121°C, 7 min), and then immediately cooled to 4°C. The first precultures of both L. bulgaricus and S. thermophilus were inoculated (1%; vol/vol) into fresh SMY media, incubated at 37°C to reach an acidity of 0.7%, and then cooled immediately to 4°C (second preculture). These second precultures were used as the inoculating seed for various milk media described below. Three S. thermophilus strains (WT and mutants) were cultured and precultured in 1/2 M17 broth supplemented with 1% lactose to determine NADH oxidase activities, DO consumption, and Eh 7 as shown in Figure 6 (A)-(E).
To examine the effect of formate or adenine supplementation on the acidification rate of S. thermophilus monocultures, 0.5 mM sodium formate (Wako Pure Chemical Industries, Ltd., Japan) or 50 µg/mL adenine hydrochloride (Wako Pure Chemical Industries, Ltd., Japan) was added to 10% skim milk medium, which contained 0.1% (wt/wt) peptides as described above. For the monoculture and coculture media, 10% (wt/ wt) skim milk was prepared by heating it to 95°C and then immediately cooling it to 43°C to prevent formate production from lactose thermal decomposition. Formate is produced from the thermal decomposition of lactose during the milk sterilization process at 99°C and 105°C (Trimigno et. al., 2020). The above treatment could minimize the formate accumulation in milk and provide insight on the behaviors of S. thermophilus and L bulgaricus and their protocooperation in milk.
The second preculture of S. thermophilus and L. bulgaricus was inoculated at 1% (vol/vol) into the skim milk medium. For L. bulgaricus and S. thermophilus mono-culture experiments, the second preculture was inoculated at 2%, and either 1 mM sodium formate (Wako Pure Chemical Industries, Ltd., Japan) or 0.1% (wt/wt) peptides generated from the digestion of casein (CE90GMM, Nippon Shinyaku Co., Ltd., Japan) was added to the skim milk medium prepared as described above. The second precultures of S. thermophilus and L. bulgaricus were then inoculated at 2% (vol/vol) into each medium. Monocultures and cocultures were incubated at 43°C.

Construction of nox and pflB knockout mutants of S. thermophilus
The nox knockout mutants of ST501 and ATCC 19258 T were obtained in the same way and the same vector was used to have ST1131 Dnox as described previously (2014, Sasaki). The nox sequence was referred to the accession numbers of CP000023.1 and it was used as an insertion fragment to construct nox knockout vector. The result confirming the construction of nox deficient strain is shown in Figure S1. (Supplementary Data; https: / / figshare .com/ s/ a31994f8b76fbb1d874a) The pflB knockout mutant was constructed as follows. The pflB sequence was referred to the accession numbers of CP000023.1 and it was used as an insertion fragment to construct pflB knockout vector.

Measurement of acidity, fermentation time, and lactate during fermentation
Acidity was measured according to the previously described method by Yamauchi et al. (2019). Fermentation time was defined as the time taken to reach 0.6% acidity with a few exceptions. The acidities used to calculate the fermentation time in each experiment are described in the Figure legends. To measure D-and L-lactate levels in the skim milk medium, protein was precipitated according to the method described by Yamauchi et al. (2019). D-and L-lactate levels were measured using high-performance liquid chromatography (HPLC) and a SUMICHIRAL OA-5000 column (Sumika Chemical Analysis Service, Ltd., Osaka, Japan) (Iwamoto, 2022).

Measurement of DO
The DO levels were measured using optical oxygen sensors (VisiFerm DO ARC 225, Hamilton, Reno, NV, USA). The DO consumption rate (µM/min) was determined using the following formula: DO consumption rate µ min

Measurement of formate concentration
Formate levels in the skim milk medium were measured using an F-kit (Roche Diagnostics K.K., Tokyo, Japan). Measurements were carried out following the manufacturer's instructions.

Measurement of NADH oxidase activity
Nox activity was measured according to the method described by Sasaki et al. (2014). One unit of activity equaled 1 µmol of NADH oxidized to NAD per min.

Measurement of redox potential
The measured redox potential (E m ) was monitored every 30 min using a portable HM-31P pH meter (DKK-TOA Corporation, Tokyo, Japan) and DM-32 P DO/pH meter (DKK-TOA Corporation) and converted to E h according to the method previously described (Tachon et al., 2010). The E h value was transformed to E h7, the redox potential at pH 7, using the Lieistner-Mirna equation: where the α constants determined for milk and casein supplemented milk were 41 and 46, respectively.

Statistical analysis
The GraphPad Prism6 software (GraphPad, San Diego, CA, USA) was used to perform statistical analysis. The statistical difference between groups was determined by 2-sided Student's t-test. One-way ANOVA (ANOVA) with post-hoc Tukey's post-hoc test was used for the comparison of more than 2 groups. The correlation between the fermentation rate and formate accumulation, and between formate accumulation and NADH oxidase activity was investigated by Pearson correlation coefficient test. p values less than 0.05 were considered statistically significant.

Monoculture fermentation time of L. bulgaricus and S. thermophilus
First, we estimated monoculture acidification rate of 6 L. bulgaricus and 7 S. thermophilus strains in the culture conditions (given below) and sorted the fast-/ slow-acidification strain as shown in Table 1.
It was fallible to assess the monoculture acidification rate of L. bulgaricus or S. thermophilus in milk without reconstitution since it was difficult to grow their monocultures normally in milk. To overcome these deficiencies, each strain was supplemented with cofactors: 1 mM sodium formate (provided by S. thermophilus in cocultures) was added to L. bulgaricus monocultures, whereas 0.1% (vol/vol) CP (peptides and amino acids that result from the digestion of casein) (provided by L. bulgaricus in cocultures) was added to S. thermophilus monocultures. Supplementing these cultures enabled the comparison of the monoculture acidification rates of L. bulgaricus and S. thermophilus.

Yogurt fermentation (coculture) rates depended on the acidification rate of S. thermophilus monoculture
We evaluated the fermentation period of cocultures of S. thermophilus ST1131 (a fast-acidifying strain) with 6 L. bulgaricus fast-and slow-acidifying strains in skim milk medium ( Figure 1C). The coculture fermentation period was almost the same when ST1131 was cultured with any of the 6 L. bulgaricus strains ( Figure  1C). As expected, the cocultures of fast-acidifying S. thermophilus ST1131 and fast-acidifying L. bulgaricus (LB2038, LB494, or LB497) showed rapid fermentation (4~6 h). Interestingly, the coculture of fast-acidifying S. thermophilus ST1131 with slow-acidifying L. bulgaricus (LB496 or LB495) also exhibited a short fermentation time. The fermentation time of LB495 was greatly shortened by coculturing with ST1131, from 21.2 to 5.7 h ( Figure 1A, C). In the case of slow-acidifying ATCC 11842 T , the coculture fermentation time was significantly longer when compared with fast-acidifying L. bulgaricus LB2038 or LB497, although it was shortened by coculture with ST1131 from 14.9 to 8.3 h ( Figure  1C). Since there was a large difference from 3.8 to 21.2 h in the monoculture fermentation time of 6 L. bulgaricus, coculture with fast-acidifying S. thermophilus ST1131reduced the time gap between them.
However, this gap reduction between the fermentation time was not observed when the fast-acidifying L. bulgaricus strain LB2038 was cultured with the 7 different S. thermophilus strains ( Figure 1D).
The fermentation time of the coculture of fast-acidifying L. bulgaricus LB2038 with slow-acidifying S. thermophilus (ST503, ST502, ATCC 19258 T or ST500) was 19-30 h, which was significantly longer than that of L. bulgaricus LB2038 with fast-acidifying S. thermophilus (ST501, ST499 or ST1131), their fermentation times were 5 -8 h. The same trend was observed when the slow-acidifying L. bulgaricus LB495 or ATCC 11842 T was fermented with the 7 strains of fast-acidifying and slow-acidifying S. thermophilus ( Figure 1D). Figure 1B (the fermentation time of S. thermophilus monocultures) and Figure 1D (the fermentation time of cocultures) showed nearly identical patterns.
The results of 24 cocultures of 7 fast-or slowacidifying S. thermophilus strains with 6 fast-or slowacidifying L. bulgaricus strains (Figures.1C, D) showed that these strains of cocultures made progress at a similar rate to that of S. thermophilus monoculture. These observations suggested that the yogurt fermentation rates correspond mainly to the acidification rate of S. thermophilus monoculture irrespective of the fast-/ slow-acidifying L. bulgaricus strains combinations considered.
Acidities in coculture were composed different llactate/d-lactate ratio depending on combination.
Here we used the content of "acidity," that is, the sum of L-lactate produced by S. thermophilus, and Dlactate produced by L. bulgaricus, and the proportion of L/D-lactate were different among combinations of their strains when their acidities were the same ( Figure 1C and D). When there was a reduction in the difference of fermentation time by coculture with fast-acidifying S. thermophilus ST1131 (Figure 1D), the ratio of L-lactate to D-lactate with fast-acidifying L. bulgaricus LB2038 was 2.9 at 0.7% acidity. However, the L/D-lactate ratio increased to 19 at 0.7% acidity, when ST1131was cocultured with a slow-acidifying L. bulgaricus strain ATCC 11842 T (Supplementary Data S3; https: / / figshare .com/ s/ a31994f8b76fbb1d874a). Even when slow-acidifying S. thermophilus strains were cocultured with fast-or slow-acidifying L. bulgaricus strains, the fermentation time was close to that of the slow-acidifying S. thermophilus. In these cases, the L/D-lactate ratio was less than 1, thereby indicating that the cell number of L. bulgaricus was larger than that of S. thermophilus.

Correlation between acidification rate and formate accumulation in 7 S. thermophilus monocultures
Our previous work showed that the S. thermophilus ST1131 nox mutant could not produce formate and its acidification was severely delayed (Sasaki, 2014). Derzelle et. al. (2005) showed that formate supply improved S. thermophilus growth in milk. These results suggested that formate is important growth stimulating factor of S. thermophilus.
Monocultures of 7 S. thermophilus strains (ST501, ST499, ST1131, ST503, ST502, ATCC 19258 T , and ST500) were performed to evaluate the correlation between the acidification rate and formate accumu- . Three L. bulgaricus strains (LB2038, LB494, and LB497) and 3 S. thermophilus strains (ST501, ST499, and ST1131) are "fast-acidifying strains." Three L. bulgaricus strains (LB496, ATCC 11842 T , and LB495) and 4 S. thermophilus strains (ST503, ST502, ATCC 19258 T , and ST500) are "slow-acidifying strains." Monocultures of L. bulgaricus or S. thermophilus were performed at 43°C in skim milk medium supplemented with 1 mM sodium formate or 0.1% peptides and amino acids that result from the digestion of casein (CP), respectively. Cocultures of L. bulgaricus with S. thermophilus were performed at 43°C in skim milk medium. Fermentation time (h) is the time required to reach 0.6% acidity, with the exception that monocultures of S. thermophilus ST500 and cocultures of L. bulgaricus ATCC 11842 T with S. thermophilus ST500 or ST503 are only 0.5% acidity because their fermentations were too slow to reach 0.6% acidity. Error bars show the SD of the results from at least 3 independent experiments. Different lowercase letters indicate significant differences in each fermentation time (P < 0.05). lation ( Figure 2). The formate accumulation of 7 S. thermophilus strains ranged from 0 to 0.65 mM and differed among strains. In the fast-acidifying strains (ST501, ST499, and ST1131), formate accumulation was 0.45, 0,64, and 0.58 mM, respectively. However, in slow-acidifying strains, ST503, ST502, ATCC 19258 T , and ST500, formate accumulation was 0.18, 0.21, 0.41, and 0 mM, respectively. The formate accumulation of the fast-acidifying strains was significantly higher than that of the slow-acidifying strains. A strong positive correlation (r = 0.838, P = 0.019) was observed between formate accumulation in milk and acidification rate in the 7 S. thermophilus strains.331-Ln333

Effect of formate supplementation on acidification of 7 S. thermophilus monocultures
The fast-acidifying strains could accumulate more than 0.45 mM formate and slow-acidifying strains accumulated lesser than 0.41 mM formate (Figure 2). Then, we examined the effect of 0.5 mM formate supplementation on the acidification rate of S. thermophilus monocultures. Monocultures of 7 S. thermophilus strains (ST501, ST499, ST1131, ST503, ST502, ATCC 19258 T , and ST500) were cultured in skim milk medium supplemented with or without 0.5 mM sodium formate to evaluate the effect of formate supplementation (Figure 3). The supplementation of 0.5 mM sodium formate significantly reduced fermentation time in 5 out of 7 strains (ST499, ST1131, ST503, ST502, and ST500). The time shortened by 0.5 mM sodium formate supplementation was 5 -33 h in the slow-acidifying strains (ST503, ST502, and ST500) and 1-2 h in the fast-acidifying strains (ST499 and ST1131). The acidification of slow-acidifying ST500 was significantly accelerated by the supplementation of sodium formate. Since we could not detect formate accumulation in ST500, we performed a sequence analysis of the pflB gene and identified a frameshift mutation at nucleotide position 471, leading to a truncated pflB protein of 191 amino acid residues. The N-terminal 156 amino acid sequences of CNRZ1066 pyruvate formate-lyase (Accession number: CP000024) were identical to those of ST500 PflB upstream of the frame shift mutation.
Although the stimulating effect of the formate supplementation on acidification rate could not be observed in ATCC 19258 T , the supplementation of 50 µg/mL adenine shortened fermentation time from 22 to 8 h. In other strains whose fermentation time were shortened by the supplementation of sodium formate (ST499, ST1131, ST503, ST502, and ST500), 50 µg/mL adenine also shortened fermentation times similarly to sodium , and ATCC 19258 T (+), ST500 (▼)) cultured at 43°C in skim milk medium supplemented with 0.1% CP. Correlation coefficient (r) was 0.838 (P = 0.019). Acidification rate (/h) was defined as the reciprocal of fermentation time which required to reach 0.6% acidity. Acidity (%) corresponds to the amount of acid produced by S. thermophilus in skim milk medium. Formate accumulation (mM) was defined as the formate concentration of the culture medium at 0.6% acidity, with the exception that ST500 was at 0.5% acidity. These values are the mean of at least 3 independent experiments. formate supplementation (data not shown). This result indicates that ATCC 19258 T may have a deficiency in the purine biosynthesis pathway due to formate and suggests that fast acidification of S. thermophilus may not only require formate, but also an intact purine biosynthesis pathway.
It was observed that a low concentration of formate (0.5 mM) had a significantly stimulating effect on S. thermophilus acidification (Figure 3), because the milk medium must be almost formate or in purine starvation condition.

Effect of Pfl deficiency on acidification of S. thermophilus ST1131
S. thermophilus ST1131, a pflB-deficient mutant (ΔpflB), was constructed to evaluate the importance of formate production in skim milk fermentation. S. thermophilus ST1131 WT and ΔpflB strains were cultured in skim milk medium supplemented with 0.1% CP (Figure 4). PflB deficiency resulted in a significant delay in fermentation time and 42 h were required for ΔpflB to reach 0.6% acidity, whereas this was achieved within 4 h for the WT of ST1131. Sodium formate supplementation shortened the fermentation time of ST1131ΔpflB to that of the WT strain (Figure 4). The results suggested that a higher amount of formate production by PFL is essential for faster acidification of S. thermophilus (Figure 2, 3, and 4).
Formate has been considered as a protocooperation factor, which S. thermophilus provide to L. bulgaricus. However, formate producing ability of S. thermophilus was a key driver to determine the acidification rate of S. thermophilus monoculture. Derzelle et al. (2005) identified formate to be a substance of formyl-tetrahydrofolate (THF) synthetase for the synthesis of THF feeding the purine biosynthetic pathway in S. thermophilus (Derzelle et al., 2005).

Formate function in the purine biosynthetic pathway and the ribonucleotides conversion in S. thermophilus
Furthermore, formate plays another prominent role as a reductant in nucleic-acid metabolism, especially in S. thermophilus. Ribonucleoside-triphosphate reductases (RNRs, EC: 1 .1 .98 .6), which are essential enzymes in DNA biosynthesis that converts all 4 ribonucleotides Yamauchi et al.: FORMATE AND NADH OXIDASE IN YOGURT FERMENTATION   Figure 3. Monoculture fermentation time of 7 S. thermophilus strains cultured at 43°C in skim milk medium supplemented with 0.1% CP without (black bar) or with 0.5 mM sodium formate (white bar). Fermentation time (h) is defined the time required to reach 0.6% acidity, with the exception that monoculture of S. thermophilus ST500 and ATCC19258 T are 0.5% acidity because their fermentations were too slow to reach 0.6% acidity. Acidity (%) corresponds to the amount of acid produced by S. thermophilus in skim milk medium. Asterisks indicate significant differences (*P < 0.05, **P < 0.01, *** P < 0.005, ****P < 0.001). Error bars show SD of the results from at least 3 independent experiments. into deoxyribonucleotides, are classified into 3 types: class I, class II, and class III. Bacteria such as E. coli, and Bacillus have class I aerobic RNRs, while S. thermophilus have class II and class III RNRs, according to the Kyoto Encyclopedia of Genes and Genomes (KEGG). The class III RNRs of S. thermophilus are anaerobic and only use formate as a reductant (Levitz et al., 2022). Therefore, formate plays a critical role in DNA replication and repair of S. thermophilus. It was suggested that the formate-producing ability was involved in nucleic-acid metabolism and corelated with the proliferation and acidification rates of S. thermophilus.
Additionally, KEGG showed that 4 strains of L. bulgaricus have anaerobic ribonucleoside-triphosphate reductase, class III RNRs which require formate as a reductant, (Ldb0211, LBU0165, LBUL0185, LD-BND0202) thereby showings that formate is an essential reductant for the DNA metabolism of L. bulgaricus.

Correlation between NADH oxidase activity and formate accumulation in 7 S. thermophilus monocultures
We previously found that Nox is required for formate production by S. thermophilus ST1131 in skim milk medium (Sasaki et al., 2014). Accordingly, the correlation between formate accumulation and NADH oxidase activities of 6 S. thermophilus strains (ST501, ST499, ST1131, ST503, ST502, and ATCC 19258 T ) was analyzed ( Figure 5). Nox activities of 6 S. thermophilus strains ranged from 0.1 to 1.0 U/mg protein, and there was a strong positive correlation (r = 0.845, P = 0.034) between Nox activity and formate accumulation ( Figure 5). This result suggested that NADH oxidase may be a factor that determines the amount of formate production in S. thermophilus.

Effect of Nox deficiency on DO consumption rate and redox potential of S. thermophilus monocultures
Although S. thermophilus ST1131 showed high Nox activity and formate accumulation, we selected 2 other strains which showed lower Nox activity and formate accumulation. We generated Nox deficient mutants (Δnox) for 2 S. thermophilus strains (ST501 and ATCC 19258 T ) based on the method of ST1131Δnox construction described by Sasaki et al. (2014) and confirmed that NADH oxidase activities were mostly inhibited not only in ST1131Δnox but also in ST501Δnox and ATCC 19258 T Δnox ( Figure 6A). The WT and the Δnox of 3 strains (ST1131, ST501, and ATCC 19258 T ) were cultured in skim milk medium supplemented with 0.1% CP to evaluate the effect of Nox deficiency on  Pfl-deficient (Δpfl) cultured at 43°C in skim milk medium supplemented with 0.1% CP without or with 0.5 mM sodium formate. Fermentation time (h) is time required to reach 0.5% acidity because the fermentations of Δpfl was too slow to reach 0.6% acidity. Acidity (%) corresponds to the amount of acid produced by S. thermophilus in skim milk medium. Error bars show SD of the results from at least 3 independent experiments. Asterisks indicate significant differences (****P < 0.001).
DO consumption rate ( Figure 6B). NADH oxidase deficiency decreased DO consumption rates in ST501 and ATCC 19258 T to the same extent as in ST1131, as previously reported (Sasaki et al., 2014). Their residual DO consumption activities were 34% (ST1131), 14% (ST501), and 28% (ATCC19258 T ) ( Figure 6B). These results showed that NADH oxidase played a key role in DO consumption in S. thermophilus strains.
A major delay in the acidification rate was observed in all Δnox when compared with their WT counterparts ( Figure 6C). The fermentation time required to reach 0.6% acidity increased over 3-fold in ST1131Δnox and 8-fold in ST501Δnox when compared with the WT, and in the case of ATCC 19258 T Δnox could not reach 0.6% and the final acidity was only 0.4% after 24 h of cultivation. Additionally, these acidification delays of Δnox strains could not be recovered under an anaerobic nitrogen-saturated condition ( Figure 6C).
We measured the redox potentials of Δnox and WT strains and in ST1131 and ST501 WT strains, the E h7 decreased below 0 mV after 3 h of cultivation and subsequently reached −50 mV after 4 h ( Figures 6D).
As the E h7 in ST1131Δnox and ST501Δnox gradually decreased to around 50 mV, they could not reach below 0 mV even after 6 h of cultivation ( Figures 6D). How-ever, in S. thermophilus ATCC 19258 T WT, wherein the Nox activity is lower than that of ST1131 and ST501 ( Figure 6A), the E h7 reached around 0 mV after 6 -7 h of cultivation but not below 0 mV even after 24 h of cultivation ( Figure 6D). In ATCC 19258 T Δnox, E h7 decreased extremely slowly when compared with WT, and reached to around 100 mV after 12 h of cultivation. The E h7 values in all Δnox were unstable across multiple experiments and tended to show higher standard deviation. We hypothesized that the other plural mechanisms tried to complement NADH oxidase deficiency by reducing the redox potential, and the E h7 values in all Δnox and became unstable.
These results showed that NADH oxidase deficiency induced the decrease of DO consumption ability (Figure 6B) and resulted in severe fermentation delay even under anaerobic conditions ( Figure 6C) and inhibition, thereby reducing the redox potential ability ( Figure  6D). Additionally, it was suggested that the fermentation of S. thermophilus required not only anaerobic conditions, but also the redox potential reducing condition produced by Nox. Jeanson et al. (2009) reported that the redox potential of oxygen-free milk (saturated with nitrogen) was still positive and some of the enzymes that lower the re- , and ATCC19258 T (+))} were cultured in the skim milk with 0.1% CP at 43°C. ST500 was removed in Figure 5, as its pfl gene had a frameshift at nucleotide position 471 leading to a truncated Pfl protein. The amounts of formate accumulation (mM) were determined in the culture medium at 0.7% acidity, with the exception that ATCC19258 T are at 0.6% acidity. NADH oxidase activities were determined using the cell cultured at 37°C in M17 broth supplemented with 1% glucose with shaking. These values are average of the results from at least 3 independent experiments and the correlation coefficient (r) between Nox activity and formate accumulation was 0.845 (P = 0.034).
dox potential are required for formate accumulation in L. lactis. As Tachon et al. (2017) reported that L. lactis has a strong reducing ability and was able to decrease milk redox potential to −220 mV (E h7 ), NADH oxidase (NoxE) and the respiratory chain components (NoxA, NoxB, MenC) were identified to contribute to the reduction of redox potential. NADH oxidase is responsible for reducing redox potential primarily by removing DO from milk, while the respiratory chain components are proposed to reduce the potential by reducing oxidizing compounds other than oxygen in milk. Unlike L. lactis, S. thermophilus does not have genes consisting of the respiratory chain, thus it is not surprising if S. thermophilus Nox has dual functions: the exclusion of DO and the reduction of oxidizing compounds in milk to achieve a low redox potential that facilitates formate production and acidification.

Effect of reducing reagent supplementation on formate production of S. thermophilus
To further analyze the function of NADH oxidase, the effect of 0.1% (vol/vol) ascorbic acid supplementation as a reducing reagent was examined on acidification rate, DO consumption rate, E h7 , and the formate accumulation of ST1131 WT (Figures 7A and B) and Δnox ( Figures 7C and D). Both strains were cultured in skim milk medium supplemented with 0.1% CP.
By supplementing ST1131 WT with ascorbic acid ( Figure 7A), the time required to reach 0.6% acidity was reduced by more than 1 h, and it took 30 min lesser to reach a concentration of 0 µM DO. With ascorbic acid supplementation, the E h7 of WT could reach −50 mV approximately 2.5 h faster, and the formate concentration reached over 0.8 mM approximately 1.5 h faster than without ascorbic acid supplementation ( Figure 7B). A significant alternation was observed when ascorbic acid supplementation was applied to ST1131 Δnox, since ascorbic acid supplementation showed stimulating effects on ST1131 WT fermentation. The acidification rate in Δnox was significantly slower when compared with WT ( Figure 7A and C), the acidity of Δnox was 0.4% at 12 h, and could not reach 0.6% ( Figure C). The redox potential for first 8 h was incapable of altering Δnox, and then for the next 4 h, Eh 7 decreased by 200 mV. Formate could not be detected at all in Δnox for 12 h of cultivation ( Figure 7D) and even after 30 h of cultivation (data not shown). However, with ascorbic acid supplementation, Eh 7 began to decrease gradually for the first 4 h, reaching −50 mV after 5 h. In the treatment of ascorbic acid supplementation, formate could be detected after 6 h and gradually increased  WT (black) and Δnox (White) using cell cultured at 37°C in M17 broth supplemented with 1% glucose with shaking. (B) Dissolved oxygen consumption rate (µM/min) of ST1131, ST501 and ATCC19258 T , WT (black) and Δnox (White) in skim milk medium. DO consumption rate (µM/min) was obtained by dividing the initial DO concentration by the time which DO concentration reduced to 0 µM. (C) Acidification curves of WT and Δnox in ST1131, ST501 and ST19258 T under aerobic / anaerobic condition in skim milk medium. The redox potential (Eh 7 ) of 3 S. thermophilus monocultures ST1131 (D), ST501 (E), and ATCC19258 T (F) in skim milk medium. In Figure 6(A), 6(B), 6(D), 6(E), 6(F) experiments, preculture which cultured in 1/2 M17 broth supplemented with 1% lactose was used. Preculture in Figure 6(C) experiment, SMY was used. Main culture in (B)~(F) were incubated at 43°C in skim milk medium supplemented with 0.1% CP. Asterisks indicate significant differences (* < 0.05, ** < 0.01, *** < 0.005, **** < 0.001). Error bars show SD of the results from at least 3 independent experiments. Though an acidification curve of ST501Δnox was examined 2 times, SD couldn't show in Figure 6 (C).
It was observed that ascorbic acid supplementation increased the redox potential dropping, formate accumulation and the acidification rate even in wild type ST1131 (Figures 7A and B). The same stimulating effect of formate supplementation on the ST1131 acidification ( Figure 3) was observed similar to ascorbic acid supplementation ( Figure 7A). ST1131 Δnox was incapable of formate accumulation without ascorbic acid supplementation and its acidification delayed deeply. The supplementation of ascorbic acid enabled ST1131 Δnox to decrease the redox potential and formate accumulation after 5 h of cultivation.
This observation showed that the formate production by PflB requires not only an anaerobic environment, but also the reduction of the redox potential in skim milk medium, and NADH oxidase provides both conditions.

Formate production complex consisting of formatelyase activating enzyme, AdoMet and PflB working under low redox potential conditions
As described above, the reducing redox potential abilities of NADH oxidase mayt be an essential factor for activating PflB. Formate was synthesized by PflB, and this reaction required the formation of a specific glycyl radical on PflB by the pyruvate Formate-lyase Activating Enzyme (PFL-AE), which is also oxygensensitive (Crain and Broderick, 2014). PFL-AE is an iron-sulfur protein, and its cofactor, S-adenosyl-Lmethionine (AdoMet), was involved in the activation of PflB. PFL-AE exists largely in combination with PflB and AdoMet. A reduced [4Fe-4S] + cluster, which provides the electron, was required for the reductive cleavage of AdoMet to generate the catalytically essential glycyl radical of PflB (Broderick et al., 2019). 1% ascorbic acid, ■: without 0.1% ascorbic acid), and acidity (%) (○: with 0.1% ascorbic acid, •: without 0.1% ascorbic acid) in ST1131 WT (A) and ST1131 Δnox (C). Redox potential: Eh 7 (□: with 0.1% ascorbic acid, ■: without 0.1% ascorbic acid), and formate accumulation (○: with 0.1% ascorbic acid, •: without 0.1% ascorbic acid) in ST1131 WT (B) and ST1131 Δnox (D). ST1131 WT and ST1131 Δnox cultured at 43˚C in skim milk medium supplemented with 0.1% CP with 0.1% ascorbic acid (white and solid line) or without (black and dotted line). Only monocultures of ST1131 Δnox were supplemented with 2 mg/ml spectinomycin. Error bars show the SD of at least three independent experiments. These sequential reactions may require the reducing redox potential conditions (Bim D., 2020).

CONCLUSIONS
This study identified that the yogurt fermentation rate was mainly dependent on the acidification rate of S. thermophilus monoculture, determined by the capacity of formate production, irrespective of the combination of fast-/slow-acidifying L. bulgaricus strains. The ΔpflB and the Δnox experiments revealed that formate may play a significant role in S. thermophilus acidification and proliferation, and formate production requires not only anaerobic conditions but also the reducing redox potential conditions maintained by Nox. Although formate has been identified as an important protocooperation and a proliferative factor of L. bulgaricus heretofore, it also plays a key role in determining the acidification rate of S. thermophilus and, consequently, yogurt fermentation. From the point of view of industrial yogurt manufacturing, it is important to select S. thermophilus strains with high NADH oxidase activities that can effectively remove oxygen and reduce the redox potential of milk to produce higher amount of formate.

ACKNOWLEDGMENTS
A part of this study was financially supported by Meiji Co., Ltd.