Lactococcus lactis ssp. cremoris MG1363 and evolved strains were cultivated in M17 medium (Oxoid™) supplemented with 0.5% glucose (w/v; named GM17) at 30°C. For static fermentation conditions (ST), the cultures were statically incubated in closed, full filled tubes. For aerated conditions (AE), the cultures were incubated in Erlenmeyer flasks filled with 10% volume media, shaking at 200rpm. For respiration-permissive condition (RES), 2μg/ml heme was added to the cultures besides the settings of the aerated conditions.

The evolution was carried out in 100-ml flasks filled with 10ml GM17 media, shaking at 200rpm at 30°C. The first culture was obtained by inoculating a colony of strain MG1363 in the medium. Thereafter every 72h, a passage was made by transferring 1/100 volume of the old culture into fresh medium in a new 100-ml flask. In total 14 passages were made before evolved strains were isolated for analysis. The cultures from the 14th passage were plated, and single colonies were purified. In total two independent cultures (A and B) were successfully maintained throughout all the passages, with nine mutants screened in culture A and five mutants in culture B for vitamin K2 content (specific concentration) under static fermentation conditions. Five out of nine mutants from culture A and all five mutants from culture B showed similar (no increase or less than 30% increase) vitamin K2 content as the wild-type strain, and we selected all three mutants (Evo1 – 3) from culture A showing significant increase in vitamin K2 content for further comparative analysis with the wild-type strain.

To obtain growth curves, overnight cultures of each strain were diluted to OD₆₀₀=0.1 (600nm; path length 10mm) in 50ml fresh GM17 and subjected to desired cultivation conditions (ST, AE or RES). Samples were taken every half hour for optical density (OD) measurement at 600nm by a spectrophotometer. To determine the survival of bacteria, viable plate counts were determined at indicated time points.

An overnight culture of each strain, obtained from static cultivation, was diluted to OD₆₀₀=0.2 with fresh GM17 in a 15-ml centrifugal tube and incubated at 30°C for 1h statically. Then, 5mM hydrogen peroxide was added (t=0) and the viable plate count of bacteria was determined every 30min for 2h in total. To stop the action of hydrogen peroxide, 2μl catalase (1.3*10⁶ u/ml) was added to the sample that was examined.

An overnight culture of each strain, obtained from static cultivation, was diluted to OD₆₀₀=0.1 with fresh GM17 in a 50-ml centrifugal tube and then incubated statically at 30°C. Thereafter, the pH of the culture was measured by a pH probe every hour, for 6h in total.

Lactococcus lactis cells were cultivated in GM17 media at 30°C for 48h under static fermentation conditions and harvested for vitamin K2 analysis. Vitamin K2 was extracted from cells as described by Liu et al. (2019). Briefly, PBS washed biomass was incubated with 10mg/ml lysozyme (Sigma) at 37°C for 1h. Four volumes of extraction buffer [n-hexane and 2-propanol at a ratio of 2:1 (v/v)] was added to the lysozyme-treated bacterial suspension and vigorously mixed by vortexing. After centrifugation at 3,000×g for 10min, the (upper) organic phase was collected. Equal volume (as extraction buffer) of n-hexane was added to the remaining lower phase, and the extraction step was repeated twice. The organic phases from each sample were combined and evaporated under a flow of N₂ gas, and re-dissolved in iso-propanol. All samples were diluted in methanol and subjected to liquid chromatography–mass spectrometry (LC–MS) analysis for identification and quantification of different forms of vitamin K2.

Samples from two experiments were analyzed exactly as described in a previous study (Liu et al., 2019) by a high-performance liquid chromatography (HPLC; UFLC, Shimadzu, Japan) system coupled with a Micro-mass Quattro Ultima MS (Waters, United States). Samples from another two experiments were analyzed using a UPLC (Thermo Scientific Vanquish) coupled with MS (Thermo Q-Exactive hybrid quadrupole-Orbitrap) as follows: 1μl sample was injected into an Acquity BEH C18 column (50mm×2.1mm, 1.7μm particle, 130A, Waters), and compounds were eluted with a gradient starting from 100% methanol to methanol/isopropanol (75%/25%) in 13.7min and maintained for 5.5min before going back to 100% methanol in 1.1min. The flow rate was 0.4ml/min, and column compartment temperature was kept at 40°C. The MS system was equipped with a heated electrospray ionization (ESI) source that was set at positive ionization mode. The capillary temperature was 150°C, and the source temperature was 450°C. The sheath gas (nitrogen) was set to 40 arbitrary units, and the ion spray voltage was 4.5kV. Data were processed using software Xcalibur (Thermo Scientific, version 2.2).

Analytical standards containing MK-1 (Santa Cruz Biotechnology), MK-4 (Sigma), MK-7 (Sigma), MK-9 (Santa Cruz Biotechnology), and vitamin K1 (Sigma) were mixed in the concentration range from 1ng/ml to 3μg/ml. Vitamin K1 (phylloquinone) was added at a concentration of 150ng/ml as an internal standard in each sample. Quantities of MK-5 and MK-6 were estimated using the formulas derived from MK-4 and MK-7 calibration curves, respectively; quantities of MK-8 and MK10 were estimated using the formula derived from MK-9 calibration curve. Values were corrected based on measurement of vitamin K1 internal standards.

Strains were all cultivated under anaerobic conditions overnight in GM17 to obtain biomass for oxygen consumption analysis. For the test under respiration-permissive conditions, 2μg/ml heme was supplemented to obtain cells with functional cytochrome bd oxidase. The cells were harvested and washed in PBS once, and OD was standardized to 2 in PBS that had already been saturated with dissolved oxygen (by shaking for 30min at 180rpm in flasks). Bacterial suspension (20ml) was placed in infusion bottles (volume 25ml) with a layer of mineral oil on top to prevent exchange of gas. An oxygen microsensor (Needle-type PSt7-02, PreSens, Germany) was inserted into the bacterial suspension. The reaction was initiated by adding glucose to a concentration of 1% in the bacterial suspension. Dissolved oxygen in the bacterial suspension was followed at room temperature by a Microx 4 oxygen meter (PreSens, Germany) for 20min with an interval of 1min. The absolute value of the slope of the linear correlation between oxygen concentration and time for each measurement was taken as the oxygen consumption rate.

Lactate, formate, acetate, acetoin, and ethanol were analyzed by HPLC. The strains were incubated in GM17 media at 30°C overnight under static or aerobic conditions. Cell-free culture supernatant was obtained by pelleting the cells at 17,000×g for 5min. To remove proteins, the supernatant was first mixed with 0.5 volume cold Carrez A solution [0.1M K₄Fe(CN)₆] and then mixed with 0.5 volume cold Carrez B solution (0.2M ZnSO₄), and centrifuged at 17,000×g for 5min. The supernatant was subjected to HPLC analysis as described by van Mastrigt et al. (2018). Briefly, 25μl sample was injected into HPLC system Ultimate 3,000 (Dionex, Idstein, Germany) equipped with an Aminex HPX-87H column (300×7.8mm). Compounds were eluted by 5mM sulfuric acid at a flow rate of 0.6ml/min at 40°C, identified by UV detectors at 220, 250, and 280nm and quantified by a refractive index detector Shodex RI-101 (Showa Denko K.K., Tokyo, Japan). The quantity of metabolites detected in un-inoculated GM17 medium was used as the initial level to calculate the production/consumption of bacterial metabolites.

The original strain MG1363 and three selected evolved strains, Evo1, Evo2, and Evo3, were subjected to genome sequencing. For genomic DNA isolation, cells collected from 1ml of an overnight culture (static incubation in GM17) of each strain was used. DNeasy Blood & Tissue Kit (Qiagen, Germany) was used to extract DNA according to the manufacture’s instruction. DNA was sequenced using the Illumina HiSeq Genome Sequencing System (GATC Biotech, Germany), read length 2×150bp. The sequencing reads were deposit under BioProject with accession number PRJNA765529 at National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA).

Sequenced reads from each genome were mapped to the reference sequence of L. lactis ssp. cremoris MG1363 (NCBI accession No. AM406671_1) using BWA (Li and Durbin, 2009) with default parameters. In all strains, more than 99.9% of all reads were mapped. The removal of PCR duplicates was carried out using Picard (Picard Toolkit, 2019). In the four sequenced strains, 935–1,141×coverage depths were reached. Variant analysis was performed using GATK’s Haplotype Caller (McKenna et al., 2010; DePristo et al., 2011) to identify single nucleotide polymorphism (SNP) and insertion and deletion (InDel).

Strains were cultivated in GM17 media at 30°C overnight (16h) under desired conditions (ST, AE, and RES). Cells were harvested from 1ml cultures, washed and resuspended in 100mM Tris (pH 8), and lysed by a sonication probe twice for 45s. For each strain and condition combination, samples were collected from three independent experiments. For all samples, 40μg proteins were used for sample preparation and analysis. Sample preparation followed the filter-assisted sample preparation protocol (FASP; Wiśniewski et al., 2009). In brief, proteins were reduced with 15mM dithiothreitol, alkylated with 20mM acrylamide, and digested with trypsin. Maximally 5μl prepared sample was injected into a 0.10×250mm ReproSil-Pur 120 C18-AQ 1.9μm beads analytical column (prepared in-house) at a constant pressure of 825bar using a 1-h gradient from 9 to 34% acetonitrile in water with 0.1% formic acid in 50min by a nanoLC-MS/MS (Thermo nLC1000 coupled to a Q Exactive-HFX). MS and MSMS AGC targets were set to 3.106, 50,000, respectively, or maximum ion injection times of 50ms (MS) and 25ms (MSMS) were used. HCD-fragmented (isolation width 1.2m/z, 24% normalized collision energy) MSMS scans of the 25 most abundant 2–5+ charged peaks in the MS scan were recorded in data-dependent mode (threshold 1.2e5, 15s exclusion duration for the selected m/z +/− 10ppm).

The MaxQuant quantitative proteomics software package was used to analyze LCMS data with all MS/MS spectra as described previously (Cox et al., 2014), and the proteome of L. lactis MG1363 (UniProt ID UP000000364) was used as the protein database. Perseus was employed for filtering and further bioinformatics and statistical analysis of the MaxQuant ProteinGroups files (Tyanova et al., 2016). Reverse hits were removed; identified protein groups should contain minimally two peptides, of which at least one is unique and one unmodified. The label-free quantitation (LFQ) intensity values were used for t test; protein groups differing by a factor of 2 or more (log fold change≤−0.3 or≥0.3) and a p-value of 0.05 or less (−log value of p≥1.3) were considered significantly different. The quality of nLC-MS/MS system was checked with PTXQC using the MaxQuant result files (Bielow et al., 2016). The protein lists and LFQ intensity values of all samples can be found in Supplementary Data Sheet. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Vizcaíno et al., 2016) partner repository with the dataset identifier PXD028721.

Where applicable, statistical significance analysis was performed in JASP (0.11.1; Love et al., 2019) using ANOVA unless specified otherwise. Post hoc multiple comparisons were conducted using Tukey’s test (two-sided), and in all cases the control group was L. lactis MG1363 (^*p≤0.05).

After sequential propagation of a culture of L. lactis ssp. cremoris MG1363 for 100 generations under intensively aerated conditions (shaking at 200rpm in flasks with 10 x headspace of the culture) for 72h in glucose supplemented M17 (GM17) media, we isolated three strains and examined them closely. All three strains (Evo1, Evo2, and Evo3) showed considerably better survival compared to the original strain MG1363 under aerated conditions, as reflected by viable plate count: When all strains were cultivated in GM17 under the aerated conditions employed in the evolution experiment for 72h, the culturability was tested at 24, 48, and 72h (Figure 1A). While the viable plate count of strain MG1363 dropped to 10⁵CFU/ml after 48h in aerated conditions from the initial 10⁹CFU/ml (not shown), the three evolved strains all maintained at 10⁹CFU/ml. After 72-h aerated cultivation, viable plate count of strain MG1363 further dropped to 10²–10³CFU/ml, while strains Evo1, Evo2, and Evo3 were still at 10⁶, 10⁹, and 10⁴CFU/ml, respectively.

We also examined the vitamin K2 content in three evolved strains in comparison to the original strain under static fermentation conditions, as this is the most relevant condition for applications. The total vitamin K2 content in MG1363 was 80±19nmol/g cell dry weight. The three evolved strains showed 50–110% higher average total vitamin K2 content than the original strain (Figure 1B). The relative abundance of MK forms produced by evolved strains are similar to MG1363: 65–70% MK-9, 20% MK-8, 5–9% MK-3, and minor amount of other forms including MK-10, MK-5 to MK-7 (Figure 1C).

As the evolved strains showed favorable traits mentioned above, we further examined the growth and survival of evolved strains under various conditions, namely static, aerated, and respiration-permissive conditions in GM17 media (Figure 2). Under static and aerated conditions, the evolved strains behaved similar to MG1363: They all reached stationary phase (around OD 3.0–3.5) in 5h, and Evo3 showed slightly lower turbidity (about 0.5 lower in OD) than the other strains when reaching stationary phase under static cultivation. Under respiration-permissive condition, MG1363, Evo1, and Evo2 all reached OD values of 4.5–5.0 in 5h, while Evo3 only reached OD of 3.0 at stationary phase.

Besides survival under aerated conditions that was used for the selection of evolved strains (Figure 1A), the survival (reflected by viable plate counts) of evolved strains and MG1363 was also tested under static and respiration-permissive conditions for 72h (Figure 3).

Under static condition, the viable counts of MG1363 dropped to 10⁷CFU/ml at 48h from the initial 10⁹CFU/ml and further decreased to 10⁴CFU/ml at 72h. All three evolved strains showed similar decrease along time, but were constantly 0.5–2 log CFU/ml higher than MG1363 at 48 and 72h (p<0.05 for Evo2 and Evo3). Under respiration-permissive condition, all strains maintained high viable counts at 10⁹CFU/ml throughout the 72h.

As the evolved strains were obtained from aerated conditions, oxygen consumption by each strain was a relevant phenotype. The oxygen consumption rates in evolved strains and MG1363 were examined under aerobic and respiration-permissive conditions (Figure 4). The biomass was obtained by cultivating each strain in GM17 media overnight anaerobically (heme added to the biomass for respiration-permissive condition), washed and resuspended in air-saturated PBS for the oxygen consumption test. Under aerobic condition, MG1363, Evo1, and Evo2 showed similar oxygen consumption rate of about 70nmol/min/OD unit, while Evo3 showed a significantly lower rate of 40nmol/min/OD unit. Under respiration-permissive condition, the oxygen consumption rates of all strains were doubled compared to the values found under aerobic condition: MG1363, Evo1, and Evo2 reached 140–160nmol/min/OD unit, while Evo3 reached 80nmol/min/OD unit.

In addition to the previous observation that the evolved strains maintained high survival under aerated cultivation conditions, we also tested the strains for resistance to oxidative stress caused by H₂O₂. Cells from early log phase (OD₆₀₀=0.2) from each strain were exposed to 5mM H₂O₂ in GM17 media, and the viable plate count was monitored for 2h (Figure 5). The viable plate count of MG1363 kept decreasing after 0.5-h exposure to H₂O₂, and dropped to 10⁴CFU/ml after 2h. In contrast, all evolved strains maintained viable counts of 10⁹CFU/ml throughout the 2-h exposure to H₂O₂.

As the acidification capacity is an important feature for lactic acid bacteria as starter cultures, we also examined the evolved strains for their ability to lower the pH in GM17 media. Cells from each strain were inoculated at OD of 0.1 in GM17 media and the pH was followed for 6h. MG1363, Evo1, and Evo2 lowered the pH from 7 to below 6 in 4h, while Evo3 showed 1-h delay than the other strains in reducing the pH to lower than 6 (Figure 6).

The primary metabolites were examined in the overnight cultures of original and evolved strains under static and aerobic condition in GM17 media. Among all tested metabolites, namely lactate, acetate, formate, acetoin, and ethanol, significant differences between strain MG1363 and evolved strains were observed for lactate and acetate under aerobic condition (Figure 7). Under anaerobic condition, all strains produced about 32mM lactate and 0.5–1mM acetate. Under aerobic condition, MG1363 produced 25mM lactate and 6.5mM acetate, Evo3 produced the same amount of lactate as MG1363 and about 1mM more acetate, while Evo1 and Evo2 produced 2.5mM less lactate but 2mM more acetate than MG1363.

Whole genome sequencing revealed the mutations in evolved strains comparing to strain MG1363 (Table 1). All three evolved strains contained SNPs in genes ldh (llmg_1120), ps435 (llmg_2107), and gapB (llmg_2539) encoding proteins L-lactate dehydrogenase, hypothetical protein, and glyceraldehyde 3-phosphate dehydrogenase, respectively. Evo1 and Evo2 shared a SNP in llmg_0907 encoding an YlbN-like protein. Evo3 contained two unique mutations: a SNP in ftsL (llmg_1680) encoding a cell division protein and a deletion in purR (llmg_2551) encoding a pur operon repressor. A few more unique mutations in Evo2 and Evo3 were revealed, but these were in pseudogenes or noncoding sequences; except for a SNP in Evo2 that could be in the promoter region of rplA (llmg_2276), the rest were not expected to have regulatory effects on gene expression. No mutations were identified in genes that are predicted to be involved in vitamin K2 biosynthesis in any of the evolved strains.

The proteomes of evolved strains and strain MG1363 were examined when cells were cultivated under static, aerobic and respiration-permissive conditions in GM17 media overnight. In total 1,387 proteins were quantified by the proteomics analysis, out of the 2,383 proteins predicted for MG1363 proteome (Uniprot proteome ID UP000000364).

Among the three tested cultivation conditions, all three evolved strains showed the highest numbers of proteins with significantly different production level (cut-off p≤0.05, fold change≥2) compared to MG1363 under aerobic conditions (Table 2). Evo1 and Evo2 had 17 and 16 proteins differently produced compared to MG1363 under the aerobic condition, while under static and respiration-permissive condition only 2–3 proteins were identified to be differently produced. Evo3 showed the biggest proteome change compared to MG1363, and the numbers of differentially produced proteins under static, aerobic, and respiration-permissive conditions were 234, 300, and 292, respectively. Under all tested conditions, a significant number of the differentially produced proteins in Evo3 compared to MG1363 are identified to be ribosomal proteins and proteins involved in purine nucleotide and amino acids biosynthesis (Supplementary Figures S1, S2).

Under the aerobic condition, three proteins were identified to be significantly (p≤0.05, fold change≥2) overproduced in all evolved strains compared to MG1363 (Supplementary Figure S3): glyceraldehyde 3-phosphate dehydrogenase (A2RIN9), universal stress protein A2 (A2RK64) and formamidopyrimidine-DNA glycosylase (A2RI84; Table 3). Quantities of these three proteins in all strains under different cultivation conditions were examined closely (Figure 8). Under static condition, the levels of the three proteins in Evo1 and Evo2 were similar to that in MG1363, while Evo3 showed doubled amount compared to the rest. Under aerated conditions, the levels of the three proteins in the evolved strains were in general 2–4 times higher than in MG1363. Under respiration-permissive condition, the amounts of the three proteins in MG1363 were close to the evolved stains, about 2–4 times higher than that of MG1363 under aerobic conditions.

Proteomics analysis did not reveal many cases of differential production of proteins encoded by genes with mutations in evolved strains compared to MG1363 in all cultivation conditions (Supplementary Table S1), with the exception that purR encoded protein was at least 10 times less in Evo3 than the other strains in all conditions. Moreover, we also did not observe significant changes in proteins that are predicted to be involved in the vitamin K2 biosynthesis pathway among all strains in the three tested cultivation conditions: static, aerobic, and respiration-permissive condition (Supplementary Table S2).