In this study, we developed a mechanism-assisted data-driven model to regulate substrate feedback to improve the production efficiency of sophorolipids (SLs). First, we used a variety of on-line biosensors to establish a multi-scale parameter detection system. We found that the production of SLs by fed-batch fermentation could be divided into three stages: a stage that was limited by cell production capacity, a stage that was inhibited by high product concentration, and a stage that was limited by oxygen supply. Subsequently, we used process parameters to develop a data-driven model, and this was then combined with the analysis of cell metabolic mechanisms. The optimal production of SLs was achieved in the first and second stages by the precise feedback regulation of substrate feeding, which increased the titer of SLs by 4.9%. The control error of the substrate was reduced from more than 15% to less than 5%. The mechanism-assisted data-driven model was then applied for semi-continuous fermentation during the production of SLs. This effectively alleviated the oxygen limitation during the third stage, and further increased the productivity of SLs to 2.30 g/L/h, 40.2% higher than the fed-batch fermentation method.

Limitations in mixing and mass transfer coupled with high hydrostatic pressures lead to significant spatial variations in dissolved oxygen (DO) concentrations in large-scale bioreactors. While traveling through different zones in the bioreactor, microbes are subjected to fluctuating DO conditions at the timescales of global circulation time. In this study, to mimic industrial-scale spatial DOT gradients, we present a scale-down model based on dynamic feast/famine regime (150 s) that leads to repetitive cycles with rapid changes in DO availability in glucose-limited chemostat cultures of Penicillium chrysogenum. The results revealed that the exposure time to the low DO level (less than 10%) imposed a significant impact on the biomass growth and penicillin productivity was considerably reduced by a factor of two, while the averaged substrate consumption rates were comparable under the DO oscillation condition compared to that of 60% DO steady-state condition. Quantitative metabolomics data showed that the DO feast/famine induced a stable and repetitive pattern with a reproducible metabolic response in time. The dynamic response of intracellular metabolites under such DO oscillating conditions showed specific differences in comparison to repetitive substrate pulse experiments. Due to invariable the specific glucose uptake rate ( q s ) during a cycle, the variation in the intracellular pools size of amino acids, sugar phosphates and organic acids was less pronounced in terms of both coverage and magnitude under DO fluctuations than under repetitive substrate pulses featured with a marked variation in the q s . Remarkably, intracellular sugar polyols were considerably increased as the hallmark metabolites to reserve carbon source and reducing equivalent, which likely provide short-term benefits in such changing environments. Furthermore, the calculated cytosolic NADH/NAD + ratio under the DO oscillating condition indicated a dynamic and higher redox state of the cytosol, which has been reported to negatively affect the maintenance of penicillin productivity. Despite the increased availability of NADPH for penicillin production under the oscillatory DO conditions, this positive effect may be counteracted by the decreased ATP supply. From an economical point of view, it is interesting to note that not only the penicillin productivity was reduced under such oscillating DO conditions, but also that of the unrecyclable byproduct ortho-hydroxyphenyl acetic acid (ο-OH-PAA) and degeneration of penicillin productivity induced by low extracellular glucose sensing. Furthermore, dynamic metabolic flux analysis based on constraining time-resolved metabolite data into genome-scale metabolic model showed that Penicillium chrysogenum metabolism shifted from penicillin production to maintaining biomass growth upon a reduction of oxygen supply. The relative decreasing fluxes of amino acid metabolic pathways and fatty acid biosynthetic pathways were assumed to relieve the energy demand for balanced cellular metabolism. Taken together, the metabolic responses of Penicillium chrysogenum to DOT gradients reported here are important for elucidating metabolic regulation mechanisms, improving bioreactor design and scale-up procedures as well as for constructing robust cell strains to cope with heterogenous industrial culture conditions.

As a novel protein post-translational modification, lysine succinylation is widely involved in metabolism regulation. In this study, we focused on the distribution of lysine succinylation sites and their physiological functions in Saccharopolyspora erythraea. Using high-resolution 4D label free mass spectrometry, a large and global protein succinylome was identified in a hypersuccinylated strain E3ΔsucC. The results showed that succinylated proteins are predominantly involved in protein synthesis-related pathways (e.g., ribosomes, tRNA) and metabolic pathways, such as the TCA cycle. Proteins in these pathways generally have a higher lysine content, suggesting that lysine succinylation may have a greater regulatory role in biochemical reactions involving acidic substrates. Motif analysis revealed that charged amino acids (D, E, K, R and W) display a more regular distribution around acylation sites, implying that the polar effect between residues may be the key factor influencing lysine succinylation. Based on predicted protein structures, we highlighted the potential impact of lysine succinylation on enzyme activity in the TCA cycle. In conclusion, this study offers valuable insights into the regulation of lysine succinylation and contributes to a comprehensive understanding of its physiological functions in actinomyces.

The ‘design-build-test-learn’ (DBTL) cycle has been adopted in rational high-throughput screening for obtaining high-yield industrial strains. However, the mismatch between build and test slows the DBTL cycle due to the lack of high-throughput analytical technologies. In this study, a highly-efficient, accurate, and non-invasive detection method of gentamicin (GM) was developed, which can provide timely feedback for the high-throughput screening of high-yield strains. Firstly, a self-made tool was established to obtain datasets in 24-well ​based on the coloring of cells. Subsequently, the random forest (RF) algorithm was found to have the highest prediction accuracy with 98.5% for the training and 91.3% for verification. Finally, a stable genetic high-yield strain (998U/mL) was successfully screened out in 3005 mutants, which was verified to improve the titer by 72.7% in a 5 L bioreactor. Moreover, the verified new datasets were updated to the model database in order to improve learning ability of DBTL cycle.

Omics approaches have been applied to understand the boosted productivity of natural products by industrial high-producing microorganisms. Here, with the updated genome sequence and transcriptomic profiles derived from high-throughput sequencing, we exploited comparative omics analysis to further enhance the biosynthesis of erythromycin in an industrial overproducer, Saccharopolyspora erythraea HL3168 E3. By comparing the genome of E3 with the wild type NRRL23338, we identified fragment deletions inside 56 coding sequences and 255 single nucleotide polymorphisms over the genome of E3. Substantial numbers of genomic variations were observed in genes responsible for pathways which were interconnected to the biosynthesis of erythromycin by supplying precursors/cofactors or by signal transduction. Through comparative transcriptomic analysis, L-glutamine/L-glutamate and 2-oxoglutarate were identified as reporter metabolites. Around the node of 2-oxoglutarate, genomic mutations were also observed. Furthermore, the transcriptomic data suggested that genes involved in the biosynthesis of erythromycin were significantly up-regulated constantly, whereas some genes in biosynthesis clusters of other secondary metabolites contained nonsense mutations and were expressed at extremely low levels. Based on the omics association analysis, readily available strategies were proposed to engineer E3 by simultaneously overexpressing sucB (coding for 2-oxoglutarate dehydrogenase E2 component) and sucA (coding for 2-oxoglutarate dehydrogenase E1 component), which increased the erythromycin titer by 71% compared to E3 in batch culture. This work provides more promising molecular targets to engineer for enhanced production of erythromycin by the overproducer.

Sophorolipids (SLs) are regarded as one of the most promising biosurfactants. However, high production costs are the main obstacle to extended SLs application. Semi-continuous fermentation, which is based on in-situ separation, is a promising technology for achieving high SLs productivity. In this study, the sedimentation mechanism of SLs was analyzed. The formation of a hydrophobic mixture of SLs and rapeseed oil was a key factor in sedimentation. And the hydrophobicity and density of the mixture determined SLs sedimentation rate. On this basis, ultrasonic enhanced sedimentation technology (UEST) was introduced, by which the sedimentation rates were increased by 46.9% to 485.4% with different ratio of rapeseed oil to SLs. UEST-assisted real-time in-situ separation and semi-continuous fermentation were performed. SLs productivity and yield were 2.15 g/L/h and 0.58 g/g, respectively, simultaneously the loss ratio of cells, glucose, and rapeseed oil were significantly reduced. This study provides the new horizon for optimization of the SLs fermentation process.

Sophorolipids (SLs) are regarded as one of the most promising biosurfactants. They have a low toxicity and are easily degradable without polluting the environment. However, high production costs are the main obstacle to extended SLs application. Semi-continuous fermentation is a promising technology for achieving high SLs productivity, which is based on in-situ separation. In this study, the sedimentation mechanism of SLs was analyzed. The formation of a hydrophobic mixture of SLs and oil was a key factor in sedimentation. The hydrophobicity and density of the mixture determined SLs sedimentation rate. According to the mechanism, ultrasonic enhanced sedimentation technology (UEST) was introduced, by which the sedimentation rates were increased by 46.9% to 485.4% with different Oil/SLs ratios. UEST-assisted real-time rational in-situ separation and semi-continuous fermentation were performed. We observed that SLs productivity and yield were 2.15 g/L and 0.58 g/g, whereas the loss ratio of cells, glucose, and oil was reduced by 68.2%, 16.2%, and 65.5%, respectively. In-situ SLs separation efficiency and rate were increased by 34.5% and 26.4%, respectively. This study provides the foundation and new horizon for the optimization of the SLs fermentation process.

A well known \emph{in silico} analysis in metabolic flux analysis is the structural identifiability analysis. It comes from the fact that some enrichment measurement sets cannot uniquely elucidate all intracellular fluxes. To the best of our knowledge, the structural identifiability analysis of dynamic isotope experiments is not available in the literature. In this work, it is shown that if one measurement plan makes the dynamic isotopic fractions balance equations structurally identifiable then for any arbitrary small time interval the plan also makes the equations structurally identifiable. Based on the fact, in order to resolve the local structural identifiablity problem of the dynamic isotopic fractions balance equations approximated with piecewise affine intracellular fluxes, one should check the local structural identifiablity for the corresponding cascaded linear time invariant system at each sampling point with the approach proposed in our earlier work (Lin \emph{et al.}, Math Biosci. 2018; 300:122-12).

A well known issue in metabolic flux analysis with labelling experiments is the structural identifiability analysis. It comes from the fact that some enrichment measurement sets cannot uniquely elucidate all intracellular fluxes. To the best of our knowledge, the structural identifiability analysis of dynamic isotope experiments is not available in the literature. In this work, it is shown that if one measurement plan makes the dynamic isotopic fractions balance equations structurally identifiable then for any arbitrary small time interval the plan also makes the equations structurally identifiable. Based on the fact, in order to resolve the local structural identifiablity problem of the dynamic isotopic fractions balance equations approximated with piecewise affine intracellular fluxes, one should check the local structural identifiablity for the corresponding cascaded linear time invariant system at each sampling point with the approach proposed in our earlier work (Lin \emph{et al.}, Math Biosci. 2018; 300:122-129).

Propanol have been widely used as a precursor for erythromycin synthesis in industrial production. However, the knowledge on the exact metabolic fate of propanol was still unclear. In the present study, the metabolic fate of propanol in industrial erythromycin-producing strain S. erythraea E3 was explored via 13C labeling experiments. An unexpected pathway in which propanol was channeled into tricarboxylic acid cycle was uncovered, resulting in uneconomic catabolism of propanol. By deleting the sucC gene, which encodes succinyl-CoA synthetase that catalyse a reaction in the unexpected propanol utilization pathway, a novel strain E3-ΔsucC was constructed. The strain E3-ΔsucC showed a significant enhancement in erythromycin production in the chemically defined medium compared to E3 (786.61 vs 392.94 mg/L). Isotopic dilution mass spectrometry metabolomics and isotopically nonstationary 13C metabolic flux analysis were employed to characterize the metabolic differences between S. erythraea E3 and E3-ΔsucC. The results showed that compared with the starting strain E3, the fluxes of pentose phosphate pathway in E3-△sucC increased by almost 200%. The most significant difference located in the tricarboxylic acid cycle was also found. The flux of the metabolic reaction catalyzed by succinyl-CoA synthetase in E3-ΔsucC was almost zero, while the glyoxylate bypass flux significantly increased. These new insights into the precursor utilization of antibiotic biosynthesis by rational metabolic engineering in S. erythraea provide the new vision in increasing industrial production of secondary metabolites.