Reorganization of a synthetic microbial consortium for one-step vitamin C fermentation
Because of the survival traits of K. vulgare, the companion bacterium is not confined to Bacillus spp., some other bacteria, such as Xanthomonas maltophilia [22], can also be a good partner. Considering these conditions, Bacillus spp. was removed in this study and a synthetic consortium of G. oxydans–K. vulgare producing 2-KGA directly from d-sorbitol was constructed (Additional file 1: Figure S2a). The relationship within this synthetic microbial consortium was analyzed for further study. The whole fermentation process of G. oxydans is divided into two stages on the basis of d-sorbitol assimilation. In the co-culture system, the relationship turns from the commensalism of the first stage to the competition of the second stage, in accordance with the d-sorbitol consumption by G. oxydans (Fig. 1a–c).
The titer of 2-KGA by this consortium was only 12.9 g/L within 36 h, and the yield was 15.0 %, which was much lower than that of the industrial two-step fermentation process (about 90 %). In order to improve the performance of this one-step fermentation process, many optimization attempts have been made, including modification of the inoculation ratio, agitation speed, and aeration rate. In this way, the titer of 2-KGA reached to 59.1 g/L within 28 h, which shorten the fermentation time by about 40 % (Additional file 1: Figure S2b). Whereas in the control experiment in which only K. vulgare (Fig. 1d) or G. oxydans (data not shown) was cultured, no 2-KGA was produced. These results showed that l-sorbose produced by G. oxydans diffused into K. vulgare cells and was subsequently oxidized. However, though the optimization of fermentation conditions indeed improved the titer and yield of 2-KGA, the natural limitation of l-sorbose consumption by G. oxydans in this consortium cannot be overcome without genetic modification. In this study, G. oxydans was cultivated in d-sorbitol seed culture medium and the composition of the culture broth from at different time points during the fermentation was measured by HPLC. We found that l-sorbose cannot be consumed until d-sorbitol was exhausted after 12 h in the mono-culture of G. oxydans, which matched the conclusion drawn by Soemphol et al. [23]. Then the accumulation of an unknown byproduct was detected while l-sorbose was consumed (Fig. 1e), which would reduce the 2-KGA production and make the efficiency too low to fully meet industrial requirements. In industrial fermentation, even one percent loss of carbon source will cause a significant financial burden. Therefore, we further optimized this two-strain consortium by alleviating the metabolic competition of G. oxydans with K. vulgare for sorbose, which was helpful for establishing a better homeostasis between microbes and making them work better together.
The relationship optimization of G. oxydans–K. vulgare consortium
The core process for the synthetic microbial consortium we designed in this study was the conversion of l-sorbose, the substrate for K. vulgare to synthesize the final product 2-KGA. After screening the genome information of G. oxydans [24], five relevant genes (Table 1) encoding FAD-dependent l-sorbose 1-dehydrogenase, NADPH-dependent l-sorbose reductase, and PTS system transporter subunit IIA in the l-sorbose consumption pathway were deleted, respectively. And five engineered G. oxydans strains, namely H1, H2, H3, H4, and H5 (Fig. 2a), were obtained with the method mentioned in “Strains” section. Fermentations of each engineered strains were carried out in flasks and jar fermentors to test the effect of the gene deletions. Compared with G. oxydans, the engineered H2, H3, H4, and H5 significantly slowed down the consumption of l-sorbose and increased the level of l-sorbose left in the broth after 30 h cultivation in l-sorbose seed culture medium in flasks (Fig. 2b). Among them, H2 and H3 were considered the most effective candidates for reducing sorbose utilization in G. oxydans. Consequently, a double mutant strain (H6) was constructed by further deletion of NADPH-dependent l-sorbose reductase (B932_1330) in H3 to perform better for 2-KGA production in the consortium. After that, H2, H3 and H6 were individually co-cultured with K. vulgare, forming consortia H2 + Kv, H3 + Kv and H6 + Kv, in the medium contained 8 % d-sorbitol as substrate. The alleviation of competition and the enhancement of mutualism were verified by the undetected byproduct (Fig. 2c), the level of remaining sorbose in the broth (Fig. 2d) and the titer of 2-KGA (Fig. 2e). As a result, H2, H3 and H6 enabled an 18.6, 15.2 and 29.6 % increase in the production of 2-KGA (70.1, 68.1 and 76.6 g/L) respectively compared to the primary consortium Go + Kv (59.1 g/L) after 28 or 36 h of cultivation. The relevant data of this study was compared with that of the conventional two-step fermentation (Additional file 1: Table S1). On one hand, the yield of 2-KGA was about 99 and 91 % for each stage with 8 % d-sorbitol as substrate in the two-step fermentation process. While in our study, it reached 89.7 % with the same amount substrate and shortened the fermentation time by about 25 %. On the other hand, our route eliminated the need for a second sterilization process, where the rate of equipment utilization can be significantly improved and the production cost can be notably saved.
The optimization of the relationship between the two microbes, G. oxydans and K. vulgare, was further studied. The consortia population compositions throughout the process were analyzed to validate the variation of the relationship. Figure 2f and g showed the relative density of different microbes in co- and mono-cultured systems. We found that the K. vulgare in the engineered consortium H2 + Kv also showed a better growth than that in the primary consortium Go + Kv (Fig. 2f), coupled with the higher production of 2-KGA in H2 + Kv. Meanwhile, the growth levels of engineered H2 and G. oxydans were similar in mono-culture. While after the introduction of K. vulgare, the engineered H2 grew much faster than the wild type since 8 h after inoculation and reached almost twice of the wild type after 28 h (Fig. 2g). In the present study, another interesting phenomenon about the initial inoculum ratio of G. oxydans to K. vulgare has been found. The inoculum ratio (%, v/v) of G. oxydans to K. vulgare was firstly set as 1:4 because of the growth defect of K. vulgare. Due to a low yield of 2-KGA, we then adjusted it to 4:1, which led to a great improvement in 2-KGA productivity (data not shown). This appears to be a counterintuitive finding that high ratio of inoculated G. oxydans was beneficial for the synthetic consortium. We speculated that because of the survival traits of K. vulgare, more G. oxydans were needed to provide more nutrients for the growth and productivity of K. vulgare. It was found that G. oxydans was the most populous consortium member throughout the whole process. However the ratio of G. oxydans to K. vulgare decreased during the fermentation in the engineered consortium, which was contrary to the original consortium Go + Kv. From this point of view, the engineered H2 promoted the growth and productivity of K. vulgare and the latter stimulated the growth of H2 in return. We hypothesized that there was more interaction of biomolecules or information signals between the two microbes in this mutualistic G. oxydans–K. vulgare consortium (Fig. 2h) compared with the primary competitive consortium. Hence, metabolomic analysis of the different consortium should be done for a comprehensive description of the relationship optimization between the members.
Metabolomic analysis on the relationship optimization of G. oxydan
s–K. vulgare consortium
Simplifying the sorbose metabolic pathway will affect not only itself alone, but also other related metabolic characteristics. Thus, the metabolome of the engineered consortium H2 + Kv was compared with the primary consortium Go + Kv to better understand the metabolic changes. It was found by PCA that the metabolomic data of the consortia Go + Kv and H2 + Kv at different sampling times (4, 8, 14, 21, 28 h) grouped clearly, respectively. An interesting phenomenon was that the metabolism of consortia Go + Kv and H2 + Kv had opposite trajectories over time (Fig. 3a, b). It indicated that the metabolic characteristics of the consortium changed after the G. oxydans was replaced by the engineered H2. Pathway enrichment analysis was then carried out on the metabolomic data of consortia Go + Kv and H2 + Kv (Fig. 3c, d), we found that the glycine, serine and threonine metabolism pathways and the pyruvate metabolism pathway were among the most significantly impacted in both Go + Kv and H2 + Kv consortia. Besides, glycerophospholipid and glycerolipid metabolism were also demonstrated significant change in the H2 + Kv consortium compared to Go + Kv. Furthermore, the metabolism significantly impacted in two consortia with different metabolic trajectories indicated that the relationship between the two strains in both consortia were different.
Improved amino acids metabolism in G. oxydans–K. vulgare consortium
A complete understanding of microbial metabolism should extend from the properties of individual strain in pure culture to the combinatorial interactions supported by complex communities. Metabolic levels of G. oxydans in monoculture were compared with those in the consortium of G. oxydans and K. vulgare to better understand the metabolic effects by K. vulgare, and the metabolic interaction between G. oxydans and K. vulgare in the synthetic microbial consortium. Our metabolomics analysis showed that the metabolism of the TCA cycle, amino acids, purines and free fatty acids were all significantly affected by the introduction of K. vulgare to the fermentation of sorbitol by G. oxydans or H2. The variations of these metabolites of the consortium were compared with those of Go or H2, and the fold changes of metabolites in the consortium after engineering relative to those in primary one are shown in Fig. 3e. In this study, when part of G. oxydans was replaced by K. vulgare for fermentation, the levels of most intracellular amino acids were found to change a lot. It was reported by Liu et al. [25] that the genes contributing to the de novo biosynthesis of histidine (His), glycine (Gly), lysine (Lys), proline (Pro), threonine (Thr), methionine (Met), leucine (Leu), and isoleucine (Ile) were absent in K. vulgare. Thus, it was supposed that the amino acids levels in the consortium would be lower than that in G. oxydans monoculture. However, five of the eight deficient amino acids including His, Pro, Thr, Leu and Ile in the consortium represented higher levels in synthetic consortium, suggesting that G. oxydans synthetized more of these amino acids, allowing for the better growth and production of K. vulgare. As a consequence, we speculated that the proper supplement of these amino acids would promote better growth and production of K. vulgare, which improved the interaction of two strains during fermentation. In order to prove this hypothesis, we investigated the effect of these amino acids on the productivity of the consortium. These five amino acids were added into the fermentation medium individually and they did enhance the ability of 2-KGA productivity to some extent as expected (Fig. 4a). Next, a mixture of His, Pro, Thr, Leu and Ile, with a final concentration of 0.7, 0.3, 0.5, 0.1, 0.5 g/L, respectively, was added to the consortium of G. oxydans and K. vulgare. With the addition of these amino acids, the yield of 2-KGA in flask cultures after 36 h of cultivation reached 88.3 %, enjoying a 41.8 % increase compared to the original consortium (62.3 %) with no addition of amino acids (Fig. 4a). In addition, the transcriptional expression level of histidinol-phosphatase, 1-pyrroline-5-carboxylate reductase, homoserine kinase and 3-isopropylmalate dehydratase in co- and mono-cultured systems at different sampling time points was evaluated by qPCR for further evidence. It was found that 1-pyrroline-5-carboxylate reductase, homoserine kinase and 3-isopropylmalate dehydratase in synthetic consortium enjoyed a similar tendency (Fig. 4b–d). The transcriptional expression of these genes in H2 + Kv consortium showed a highest level at the early point of exponential phase (14 h) as expected, which was in accordance with the tendency of growth and productivity of K. vulgare. However, the levels of these genes were much lower in Go + Kv consortium and no expression could be detected in the mono-culture of K. vulgare. From this point of view, it proved that with the stimulation of K. vulgare, G. oxydans synthetized more of these amino acids for better growth and production of K. vulgare.
Improved purines metabolism in G. oxydans–K. vulgare consortium
In addition to amino acid biosynthesis deficiencies, K. vulgare was reported to be insufficient in purine nucleotide biosynthesis [26, 27]. Our previous study also found that supplement of purines did have certain positive effects on the cell growth and the 2-KGA productivity of K. vulgare. [12]. In this study, adenine and adenosine contents were both higher in consortia Go + Kv and H2 + Kv when compared with mono-cultured G. oxydans and H2, respectively (Fig. 3e). We also found that adenine and adenosine were undetected in mono-cultured K. vulgare. This suggests that G. oxydans provides these purines to K. vulgare when co-cultured, while in return, K. vulgare stimulates the biosynthesis of purines in G. oxydans. In addition, the levels of these purines decreased after engineering, which indicated the gene deletion did affect the biosynthesis of purines. However, the increase of the purines levels in co-cultured H2 + Kv compared to the mono-cultured H2 was larger than that in co-cultured Go + Kv comparing to G. oxydans, which suggested that co-culturing with K. vulgare promoted the biosynthesis of purines more in the engineered G. oxydans than that in the wild type. Therefore, on one hand, we suggested that G. oxydans provided substrates and nutrients for K. vulgare. On the other hand, K. vulgare gave some feedback to stimulate the synthesis of nutrients in G. oxydans (Fig. 5).
Improved fatty acids metabolism in G. oxydans–K. vulgare consortium
In this study, all the detected free fatty acids represented higher levels in consortium samples than that in mono-cultured G. oxydans samples, especially the unsaturated fatty acids including oleic acid (18:1), elaidic acid (18:1), and palmitelaidic acid (16:1). All of these three fatty acids were presented at over 20-fold higher levels in both consortia compared to monoculture. Additionally, their levels in consortium Go + Kv were higher than those in H2 + Kv, respectively (Fig. 3e). It was reported that the increased unsaturated fatty acid level facilitated the stress defense [28]. Thus, we speculated that G. oxydans might be subjected to several stresses after co-cultured with K. vulgare, such as the changed growth environment caused by the metabolites secreted by K. vulgare. More unsaturated fatty acids were synthesized by G. oxydans to respond to the pressure of the co-culture conditions. Compared to Go + Kv, the lower levels of unsaturated fatty acids in consortium H2 + Kv suggested that the engineered H2 possessed preferable adaptability to the environment co-culture with K. vulgare. On the other hand, more unsaturated fatty acids would increase the cell membrane fluidity and permeability under unfavorable conditions by affecting the plasma membrane integrity, fluidity and function [29]. The dramatic increase in the levels of these unsaturated fatty acids may indicate that cells in this consortium increased their membrane permeability for exchanging more nutrients, which would promote the interaction between two strains.