Gluconobacter oxydans WSH-004, which used D-sorbitol to produce l-sorbose and 2-KLG, was screened in a previous study (Chen et al., 2019). The D-sorbitol medium (10 g/L of yeast extract and 80 g/L of D-sorbitol) was used for culturing the G. oxydans strains. In agar plates, 2% agar was added. Adaptive evolution of a D-sorbitol medium (10 g/L of yeast extract and 300 g/L of D-sorbitol) was used for adaptive evolution. Carrier oil (mineral oil with 12 g/L of Span 80; Sigma–Aldrich, Germany) and MMC chips for MMC were purchased from Wuxi Tmaxtree Biotechnology Co., Ltd. (Wuxi, China).

G. oxydans WSH-004 preserved in glycerol tubes was inoculated to solid D-sorbitol medium plates and cultured for 48 h at 30°C. A single colony was inoculated into 250-mL flasks containing 25 mL of D-sorbitol medium for 48 h at 30°C and 220 rpm. The seed was injected into a special sterilized seed bottle (Figure 1, bottle 3) for adaptive evolution in MMC (Jian et al., 2020). Meanwhile, the D-sorbitol medium was injected into a special sterilized medium bottle (Figure 1, bottle 2), while the adaptive evolution D-sorbitol medium was injected into another special sterilized medium bottle (Figure 1, bottle 1). The first step for adaptive evolution in MMC was the microdroplet formation (Figure 1A), in which pumps 2 and 3 were operated to form 2 μL microdroplets. The second step involved culturing the microdroplets in MMC (Figure 1B), in which pumps 1 and 2 were operated to move the microdroplets in special tubes that could pass through the air but not through water and cells. A key step for the adaptive evolution and continuous culture is the droplet segmentation fusion-generation (Figures 1C,D). After a period of adaptive evolution in MMC, the droplet with a higher OD₆₀₀ value (Figure 1F) was collected and spread on the D-sorbitol medium plates with selective pressure (high D-sorbitol concentration or/and high-temperature conditions).

G. oxydans MMC4 obtained in the last round of adaptive evolution to the high D-sorbitol concentration was used as the start strain. The single colony of G. oxydans MMC4 was inoculated into 250-ml flasks containing 25 ml of D-sorbitol medium for 48 h at 30°C and 220 rpm to get the seeds. All steps were the same (microdroplet formation, incubation, droplet segmentation fusion-generation, and sorting), except for the gradual increase in the metal bath temperature instead of increasing the D-sorbitol concentration.

For verifying the tolerance of evolved strains to different D-sorbitol concentrations and temperature, a single colony of the strain was inoculated into 250-ml flasks containing 25 ml of D-sorbitol medium for 48 h at 30°C and 220 rpm for seed culture. Then, 5 ml of the seed culture was transferred to 25 ml of the culture medium containing different D-sorbitol concentrations (50–400 g/L of D-sorbitol, 10 g/L of yeast extract, and 50 μg/ml of cefoxitin). The flasks were then cultured at 220 rpm at 30, 35, or 40°C.

A single colony of the evolved strains isolated from the evolved populations was cultivated for 36–48 h in 25 ml of D-sorbitol medium until the OD₆₀₀ value reached 3. A genomic DNA extraction kit (Sangon Biotech, Shanghai) was used to extract genomic DNA of G. oxydans strains using the protocol recommended by the manufacturer. High-throughput genome sequencing was performed with the Illumina Hiseq using 2 × 150 paired-end reads targeting a genome. The average coverage for each sample was more than 300X. Genome sequencing was performed by Sangon Biotech. Referring to the GATK best practice–recommended design process, BWA was used to compare the effective data of the sample to the reference genome (Kumar et al., 2019). The SAM tool was used to compare the results for format conversion and sorting, and statistical comparison results (Li et al., 2009). The HaplotypeCaller of GATK was employed to analyze the genetic differences between the adaptive evolution strains and wild-type strain, merge and integrate the analysis of different samples, and obtain sample variation information (McKenna et al., 2010).

Fermentation studies were performed in a 5-L bioreactor (T&J Bioengineering, Shanghai, China) with a working volume of 3 L. For batch fermentation, a single colony of the strain was inoculated into 500-ml flasks containing 50 ml of D-sorbitol medium for 48 h at 30°C and 220 rpm. Then, 300 ml of the seed culture was inoculated into a bioreactor with 2.7 L of the culture medium containing different D-sorbitol concentrations (300 g/L of D-sorbitol and 10 g/L of yeast extract, or 80 g/L of D-sorbitol and 10 g/L of yeast extract). The temperature was maintained at 30 or 37°C, the agitation speed was controlled at 400 rpm, and the aeration rate was 1 vvm.

D-sorbitol and l-sorbose in the fermentation broth were evaluated by Shimadzu high-performance liquid chromatography equipped with an Aminex HPX-87H column (Bio-Rad, Hercules, CA, United States) at 40°C with a flow rate of 0.5 ml/min and 5 mmol/L of H₂SO₄ as the eluent. A refractive index detector was used for detection (Chen et al., 2019).

G. oxydans WSH-004 was screened from the soil in a previous study (Chen et al., 2019). It directly converted D-sorbitol into 2-KLG without obvious accumulation of by-products during the fermentation process. Therefore, G. oxydans WSH-004 had the potential to be a one-step fermentation strain for the production of 2-KLG, which is the direct precursor of vitamin C. G. oxydans WSH-004 was grown with different D-sorbitol concentrations (Figure 2). The results showed that the OD₆₀₀ value of G. oxydans WSH-004 in the stationary phase could reach 2.7 in 50 g/L of D-sorbitol medium, which was similar to that of most other commonly available G. oxydans strains. When the concentration of D-sorbitol reached 300 g/L, which is the commonly used concentration for the industrial-scale production of l-sorbose from D-sorbitol (Zhou et al., 2019), the growth of G. oxydans WSH-004 was significantly inhibited with a final OD₆₀₀ value of 0.7. The results showed that the tolerance of G. oxydans WSH-004 to high D-sorbitol concentration was poor and should be significantly improved to fulfill the industrial-scale production of l-sorbose and 2-KLG.

The strain G. oxydans WSH-004 was continuously cultured in the MMC system with different D-sorbitol concentrations to improve the tolerance of the strain to higher D-sorbitol concentrations. The growth curve of G. oxydans WSH-004 was first checked in the MMC system. The result showed that the OD₆₀₀ value reached the stationary phase after 10 h with a final OD₆₀₀ value of about 1.2 and a typical growth curve in the MMC system (Figure 3A). The value was different from that in shaking flask fermentation because of the difference in both culture conditions and detection methods. The evolution of tolerance to high D-sorbitol concentration was carried out. The automatic passage time was set to 15 h. After every three passages, the D-sorbitol concentration was automatically increased according to the set program. During the growth of G. oxydans WSH-004, the OD₆₀₀ value of every droplet was recorded by MMC (Figure 3B). When the D-sorbitol concentration of G. oxydans WSH-004 reached 300 g/L, the OD₆₀₀ values of different drops were quite different. The droplets with a higher OD₆₀₀ value were harvested, cultured in 300 g/L of D-sorbitol medium at 30°C, and then diluted and spread on D-sorbitol plates containing 300 g/L of D-sorbitol. Two single colonies of evolved strains, namely G. oxydans MMC1 and MMC2, that could grow well on 300 g/L of D-sorbitol plates, were picked and verified.

G. oxydans MMC2 was selected as the starting strain for the second round of adaptive evolution to high D-sorbitol concentration in MMC to stabilize the capacity of G. oxydans to tolerate high D-sorbitol concentrations. In the second round of adaptive evolution, no significant growth difference was found among different drops. All drops could be cultured stably in MMC (Figure 3C). After 14 days of continuous culture in MMC, the droplets with higher OD₆₀₀ values were harvested, cultured in 300 g/L of D-sorbitol medium at 30°C, and then diluted and spread on D-sorbitol plates containing 300 g/L of D-sorbitol. Two single colonies of evolved strains, namely G. oxydans MMC3 and MMC4, that could grow well on 300 g/L of D-sorbitol plates, were picked and verified.

High-temperature fermentation has many advantages in industrial production. In the process of adaptive evolution to high-temperature tolerance, G. oxydans MMC4 was used as the starting strain for performing adaptive evolution. The D-sorbitol concentration and culture temperature were initially set at 300 g/L and 30°C, respectively. The result showed that the OD₆₀₀ value could reach to about 0.7 in the MMC system with 300 g/L of D-sorbitol medium at 30°C. Then, the culture temperature was gradually raised from 30 to 40°C (Figure 4), and the D-sorbitol concentration was maintained at 300 g/L. When the incubation temperature was increased to 35°C, the maximum OD₆₀₀ value reached to about 0.5 (Figure 4A). In the later adaptive evolution of temperature tolerance (the temperature was gradually increased from 35 to 39°C), the growth of G. oxydans was basically unaffected with a maximum OD₆₀₀ value of about 0.5 (Figure 4A). After 26 days of adaptive evolution in MMC, the droplets with higher OD₆₀₀ value were harvested and then cultured in 300 g/L of D-sorbitol concentration at 39°C. As strains isolated from all drops could not grow at less than 40°C, 39°C was the highest temperature that G. oxydans could withstand in this round (Figure 4A). After 27 days of continuous culture in MMC, the droplets with higher OD₆₀₀ value were harvested, cultured in 300 g/L of D-sorbitol medium at 39°C, diluted and spread on D-sorbitol plates containing 300 g/L of D-sorbitol, and cultured at 39°C. Two single colonies of evolved strains, namely G. oxydans MMC5 and MMC6, that grew well on 300 g/L of D-sorbitol plates, were picked and verified.

Taking G. oxydans MMC6 as the starting strain, the adaptive evolution was further carried out. When the culture temperature increased to 39°C, the evolved strain grew normally (Figure 4B). Then, the evolution temperature was altered between 39 and 40°C (Figure 4B). Finally, when the strains grew stably at 40°C, the droplets with higher OD₆₀₀ value were harvested, cultured in 300 g/L of D-sorbitol medium at 40°C, diluted and spread on D-sorbitol plates containing 300 g/L of D-sorbitol, and cultured at 40°C. Two single colonies of evolved strains, namely G. oxydans MMC7 and MMC8, that grew well on 300 g/L of D-sorbitol plates, were picked and verified.

Further adaptive evolution of G. oxydans MMC8 was conducted to stabilize the capacity of G. oxydans to tolerate both high D-sorbitol concentration and high temperature. The D-sorbitol concentration and culture temperature were initially set at 300 g/L and 37°C, respectively. Once the growth of cells reached the stationary phase, the adaptive evolution temperature was gradually increased from 37 to 40°C (Figure 4C). Finally, when the cells grew stably at 40°C, the droplets with higher OD₆₀₀ value were harvested, cultured in 300 g/L of D-sorbitol medium at 40°C, diluted and spread on D-sorbitol plates containing 300 g/L of D-sorbitol, and cultured at 40°C. Two single colonies of evolved strains, namely G. oxydans MMC9 and MMC10, that grew well on 300 g/L of D-sorbitol plates, were picked and verified.

Both wild-type and evolved strains were cultured under different D-sorbitol concentrations to verify the D-sorbitol tolerance of the evolved strains. In 50 g/L of D-sorbitol medium, the final OD₆₀₀ values of both G. oxydans MMC2 and MMC4 reached to about 2.9, which was higher than that of G. oxydans WSH-004 (Figure 5A). G. oxydans MMC2 and MMC4 grew better in 100 g/L of D-sorbitol medium than in 50 g/L of the medium. The final OD₆₀₀ values of G. oxydans MMC2 and MMC4 was about 3.2 (Figure 4B). In 100–300 g/L of D-sorbitol medium, G. oxydans MMC2 and MMC4 grew better than G. oxydans WSH-004 (Figures 5C–F). When the D-sorbitol concentration was increased to 350 g/L, G. oxydans WSH-004 could not grow, while G. oxydans MMC2 and MMC4 grew to a final OD₆₀₀ value of about 1.2 (Figure 5G). When D-sorbitol concentration was increased to 400 g/L, none of the strains could grow (Figure 5H).

G. oxydans MMC4-MMC10 grew better than the wild-type strain in the 50 g/L D-sorbitol media at 30 and 35°C, and the OD₆₀₀ value of the evolved strains (about 3.5) was higher than that of the wild-type strains (about 2.0) (Figures 6A-D). When the temperature was 40°C, the growth of all strains was inhibited in 50 g/L of D-sorbitol medium (Figure 6E). However, 300 g/L of D-sorbitol could benefit cell growth at 40°C. The evolved strains grew much better at 40°C than in 50 g/L of D-sorbitol medium. The final OD₆₀₀ value reached to about 0.75 (Figure 6F). Therefore, high D-sorbitol concentration could protect G. oxydans under high-temperature conditions.

Batch fermentation of G. oxydans WSH-004 and evolved strain G. oxydans MMC10 was conducted in a 5-L bioreactor to verify the production capacity of l-sorbose. Both G. oxydans WSH-004 and G. oxydans MMC10 could completely convert 80 g/L of D-sorbitol into l-sorbose at 30°C. However, the required fermentation time was quite different. It took about 60 and 25 h for G. oxydans WSH-004 and G. oxydans MMC10, respectively, to completely convert the D-sorbitol to l-sorbose, with productivities of 1.33 g/(Lh) and 3.20 g/(Lh), respectively (Figures 7A,B). When the concentration of D-sorbitol was increased to 300 g/L, the growth of G. oxydans WSH-004 was completely inhibited, while G. oxydans MMC10 could still completely convert D-sorbitol into l-sorbose in 96 h (Figures 7C,D). The productivity of G. oxydans MMC10 with the 300 g/L of D-sorbitol fermentation medium was 3.13 g/(Lh), which was similar to that in the 80 g/L of D-sorbitol fermentation medium.

Then, the fermentation temperature was raised to 37°C. G. oxydans MMC10 grew normally, while G. oxydans WSH-004 could not grow with 80 g/L or 300 g/L of D-sorbitol (Figures 7E,F). G. oxydans MMC10 could not completely convert D-sorbitol into l-sorbose at 37°C in 80 g/L of D-sorbitol medium. The final concentration of l-sorbose was 74 g/L at 35 h, with a productivity of 2.11 g/(Lh). Besides, G. oxydans MMC10 could not completely convert D-sorbitol into l-sorbose in 300 g/L of D-sorbitol medium at 37°C. However, the strain grew better than in the 80 g/L of D-sorbitol medium, and the OD₆₀₀ value reached to 2.8. The final l-sorbose concentration was 160 g/L after 60 h, with a productivity of 2.67 g/(Lh) (Figures 7G,H). It was presumed that the enzyme activity and stability of D-sorbitol dehydrogenase were affected under high-temperature conditions.

Genome sequencing was performed on evolved strains to identify mutations that might contribute to the improved phenotypes in the evolved populations. Based on the comparative genomic analysis, 104 mutations were detected among the ten evolved strains comprising 53 point mutations and 51 insertions or deletions. The genetic variations have been listed in Supplementary Material 1 and Supplementary Material 2. Functional annotations of the mutated genes (Figure 8A) showed that the mutated genes were mainly related to translation (ribosomal small subunits S17 and S3, and the major ribosomal subunits L2, L3, L6, L24, and L29), amino acid and nucleotide metabolism, carbohydrate metabolism, and energy metabolism. In the same time, all evolved strains have the ribosome mutation-related genes and this gene was listed in Table 1. A phylogenetic tree of evolved strains and wild-type strains was built based on the SNP of each sample using the adjacency NJ algorithm (Figure 8B). According to the clusterProfiler for feature enrichment analysis, mutant genes were mainly involved in ribosomes, ribosomal protein complexes, and protein mutations that constituted ribosomes (Figure 8C).