Analysis of the cell wall binding domain in bacteriocin-like lysin LysL from Lactococcus lactis LAC460.

Wild-type Lactococcus lactis strain LAC460 secretes prophage-encoded bacteriocin-like lysin LysL, which kills some Lactococcus strains, but has no lytic effect on the producer. LysL carries two N-terminal enzymatic active domains (EAD), and an unknown C-terminus without homology to known domains. This study aimed to determine whether the C-terminus of LysL carries a cell wall binding domain (CBD) for target specificity of LysL. The C-terminal putative CBD region of LysL was fused with His-tagged green fluorescent protein (HGFPuv). The HGFPuv_CBDlysL gene fusion was ligated into the pASG-IBA4 vector, and introduced into Escherichia coli. The fusion protein was produced and purified with affinity chromatography. To analyse the binding of HGFPuv_CBDLysL to Lactococcus cells, the protein was mixed with LysL-sensitive and LysL-resistant strains, including the LysL-producer LAC460, and the fluorescence of the cells was analysed. As seen in fluorescence microscope, HGFPuv_CBDLysL decorated the cell surface of LysL-sensitive L. cremoris MG1614 with green fluorescence, whereas the resistant L. lactis strains LM0230 and LAC460 remained unfluorescent. The fluorescence plate reader confirmed the microscopy results detecting fluorescence only from four tested LysL-sensitive strains but not from 11 tested LysL-resistant strains. Specific binding of HGFPuv_CBDLysL onto the LysL-sensitive cells but not onto the LysL-resistant strains indicates that the C-terminus of LysL contains specific CBD. In conclusion, this report presents experimental evidence of the presence of a CBD in a lactococcal phage lysin. Moreover, the inability of HGFPuv_CBDLysL to bind to the LysL producer LAC460 may partly explain the host's resistance to its own prophage lysin.

Method to Assess the Intracellular Fate and Bioavailability of Clusterin Using Live Cell Confocal Microscopy Imaging.

Clusterin, also known as apolipoprotein J, is an ATP-independent holdase chaperone protein. Clusterin is involved in various functions including protein quality control and lipid transport. Though clusterin is secreted upon stress, the intracellular fate of clusterin after a stress response is not well understood. The protocol described here utilizes clusterin tagged to fluorescent proteins like green fluorescent protein and red fluorescent protein to understand the intracellular fate of clusterin.

Fluorescently Tagged Verticillium dahliae to Understand the Infection Process on Cotton (Gossypium hirsutum) and Weed Plant Species.

Verticillium wilt is a soil-borne disease caused by distinct vegetative compatibility groups (VCG) of the fungus Verticillium dahliae. Defoliating (VCG 1A) and non-defoliating (VCG 2A) pathotypes of V. dahliae have contributed to yield losses of cotton production in Australia. To study the virulence and the infection process of V. dahliae on cotton, two isolates, one representing each VCG, have been transformed with fluorescent protein genes. The transformants maintained their ability to infect the host, and both strains were observed to move through the plant vasculature to induce wilt symptoms. Furthermore, virulence testing suggests that the cotton V. dahliae strains can endophytically colonise common weed plant species found in the Australian landscape, and that is contrasted by their ability to infect and colonise native tobacco plants. The fluorescently labelled strains of V. dahliae not only allowed us to gain a thorough understanding of the infection process but also provided a method to rapidly identify recovered isolates from host colonisation studies.

Expanding the Cell-Free Reporter Protein Toolbox by Employing a Split mNeonGreen System to Reduce Protein Synthesis Workload.

The cell-free system offers potential advantages in biosensor applications, but its limited time for protein synthesis poses a challenge in creating enough fluorescent signals to detect low limits of the analyte while providing a robust sensing module at the beginning. In this study, we harnessed split versions of fluorescent proteins, particularly split superfolder green fluorescent protein and mNeonGreen, to increase the number of reporter units made before the reaction ceased and enhance the detection limit in the cell-free system. A comparative analysis of the expression of 1-10 and 11th segments of beta strands in both whole-cell and cell-free platforms revealed distinct fluorescence patterns. Moreover, the integration of SynZip peptide linkers substantially improved complementation. The split protein reporter system could enable higher reporter output when sensing low analyte levels in the cell-free system, broadening the toolbox of the cell-free biosensor repertoire.

Natural-Target-Mimicking Translocation-Based Fluorescent Sensor for Detection of SARS-CoV-2 PLpro Protease Activity and Virus Infection in Living Cells.

Papain-like protease PLpro, a domain within a large polyfunctional protein, nsp3, plays key roles in the life cycle of SARS-CoV-2, being responsible for the first events of cleavage of a polyprotein into individual proteins (nsp1-4) as well as for the suppression of cellular immunity. Here, we developed a new genetically encoded fluorescent sensor, named PLpro-ERNuc, for detection of PLpro activity in living cells using a translocation-based readout. The sensor was designed as follows. A fragment of nsp3 protein was used to direct the sensor on the cytoplasmic surface of the endoplasmic reticulum (ER) membrane, thus closely mimicking the natural target of PLpro. The fluorescent part included two bright fluorescent proteins-red mScarlet I and green mNeonGreen-separated by a linker with the PLpro cleavage site. A nuclear localization signal (NLS) was attached to ensure accumulation of mNeonGreen into the nucleus upon cleavage. We tested PLpro-ERNuc in a model of recombinant PLpro expressed in HeLa cells. The sensor demonstrated the expected cytoplasmic reticular network in the red and green channels in the absence of protease, and efficient translocation of the green signal into nuclei in the PLpro-expressing cells (14-fold increase in the nucleus/cytoplasm ratio). Then, we used PLpro-ERNuc in a model of Huh7.5 cells infected with the SARS-CoV-2 virus, where it showed robust ER-to-nucleus translocation of the green signal in the infected cells 24 h post infection. We believe that PLpro-ERNuc represents a useful tool for screening PLpro inhibitors as well as for monitoring virus spread in a culture.

Viscosity-Induced Emission of 5-(Diarylmethylene)imidazolone with Extended Conjugation via Attachment of N-Methylpyrrole at the 2-Position.

We have developed a series of 2-monoaryl-5-diarylmethylene analogs of the green fluorescent protein chromophore to study their viscosity-induced emission (VIE) properties. The analogs were synthesized by a condensation with methyl imidate and N-(diarylmethylene)glycinate. Among the analogs, the N-methylpyrrol-2-yl-substituted analog 1h induced the most remarkable VIE behavior in triglyceride and lipid bilayers probably due to the high π-electron-rich property of the pyrrole ring. The pyrrole substituent in imidazolone analogs can be expected to become a common template for introducing VIE behavior.

Head-to-head Comparison of Green Fluorescent Protein (GFP) Imaging With Luciferase-luciferin Imaging In Vivo Using Single-nanometer Laser-excitation Tuning and an Ultra-low-light-detection Camera and Optics Demonstrates the Superiority of GFP.

BACKGROUND/AIM: Genetic reporters encoding fluorescent proteins or luciferase have been used in vivo for the last three decades with claims about their superiority or inferiority over each other. In the present report, a head-to-head in vivo comparison of green fluorescent protein (GFP) fluorescence imaging and luciferase-luciferin imaging, using single-nanometer laser-excitation tuning of fluorescence excitation and an ultra-low-light-detection camera and optics was performed. MATERIALS AND METHODS: Mouse Lewis-lung carcinoma cells labeled with GFP (LLC-GFP) or luciferase (LL/2-Luc2) were injected subcutaneously into the flank of nude mice. One week after injection, GFP-fluorescence imaging and luciferase-luciferin imaging was performed using the UVP Biospectrum Advanced system with excitation at 487 nm and peak emission at 513 nm for GFP, and with emission at 560 nm for luciferase-luciferin. GFP fluorescence images were obtained at 0, 10, and 20 min. Luciferase-luciferin images were obtained 10 and 20 min after the injection of D-luciferin. RESULTS: The intensity of GFP images was 55,909 at 0 min, 56,186 at 10 min, and 57,085 at 20 min, and maintained after 20 min. The intensity of luciferase-luciferin images was 28,065 at 10 min after the injection of D-luciferin and 5,199 at 20 min after the injection. The intensity of luciferase-luciferin images decreased by approximately 80% at 20 min compared to 10 min. An exposure time of 30 s for luciferase-luciferin imaging was needed compared to 100 ms for GFP fluorescence imaging in order to detect signals. CONCLUSION: An imaging system with single-nanometer tuning fluorescence excitation and an ultra-low-light detection camera and optics was able to directly visualize both GFP and luciferase-luciferin images in vivo. The intensity and stability of the signals were both greater for GFP than for luciferase-luciferin, and the exposure time for GFP was 300 times faster, demonstrating the superiority of GFP.

Comparing the biological activity and composition of Cerebrolysin with other peptide preparations.

Neurological disorders, ranging from acute forms such as stroke and traumatic brain injury to neurodegenerative diseases like dementia, are the leading cause of disability-adjusted life years (DALYs) worldwide. A promising approach to address these conditions and promote nervous system regeneration is the use of the neuropeptide preparation Cerebrolysin, which has been shown to be effective in both clinical and preclinical studies. Despite claims of similar clinical efficacy and safety by several peptide preparations, concerns regarding their generic composition and efficacy have been previously raised. Based on these reports, we analyzed the peptide composition and neurotrophic activity of several peptide preparations allegedly similar to Cerebrolysin and approved in some countries for treating neurological diseases. Our results demonstrate that these preparations lack relevant biological activity and that the peptide composition is significantly different from Cerebrolysin. peptide.

Fluorescent riboswitch-controlled biosensors for the genome scale analysis of metabolic pathways.

Fluorescent detection in cells has been tremendously developed over the years and now benefits from a large array of reporters that can provide sensitive and specific detection in real time. However, the intracellular monitoring of metabolite levels still poses great challenges due to the often complex nature of detected metabolites. Here, we provide a systematic analysis of thiamin pyrophosphate (TPP) metabolism in Escherichia coli by using a TPP-sensing riboswitch that controls the expression of the fluorescent gfp reporter. By comparing different combinations of reporter fusions and TPP-sensing riboswitches, we determine key elements that are associated with strong TPP-dependent sensing. Furthermore, by using the Keio collection as a proxy for growth conditions differing in TPP levels, we perform a high-throughput screen analysis using high-density solid agar plates. Our study reveals several genes whose deletion leads to increased or decreased TPP levels. The approach developed here could be applicable to other riboswitches and reporter genes, thus representing a framework onto which further development could lead to highly sophisticated detection platforms allowing metabolic screens and identification of orphan riboswitches.

GFP-labeled Schwann cell-like cells derived from hair follicle epidermal neural crest stem cells promote the acellular nerve allografts to repair facial nerve defects in rats.

BACKGROUND: Acellular nerve allografts (ANAs) have been successfully applied to bridge facial nerve defects, and transplantation of stem cells may enhance the regenerative results. Up to now, application of hair follicle epidermal neural crest stem cell-derived Schwann cell-like cells (EPI-NCSC-SCLCs) combined with ANAs for bridging facial nerve defects has not been reported. METHODS: The effect of ANAs laden with green fluorescent protein (GFP)-labeled EPI-NCSC-SCLCs (ANA + cells) on bridging rat facial nerve trunk defects (5-mm-long) was detected by functional and morphological examination, as compared with autografts and ANAs, respectively. RESULTS: (1) EPI-NCSC-SCLCs had good compatibility with ANAs in vitro. (2) In the ANA + cells group, the GFP signals were observed by in vivo imaging system for small animals within 8 weeks, and GFP-labeled EPI-NCSC-SCLCs were detected in the tissue slices at 16 weeks postoperatively. (3) The facial symmetry at rest after surgery in the ANA + cells group was better than that in the ANA group (p < 0.05), and similar to that in the autograft group (p > 0.05). The initial recovery time of vibrissal and eyelid movement in the ANA group was 2 weeks later than that in the other two groups. (4) The myelinated fibers, myelin sheath thickness and diameter of the axons of the buccal branches in the ANA group were significantly worse than those in the other two groups (P < 0.05), and the results in the ANA + cells group were similar to those in the autograft group (p > 0.05). CONCLUSIONS: EPI-NCSC-SCLCs could promote functional and morphological recovery of rat facial nerve defects, and GFP labeling could track the transplanted EPI-NCSC-SCLCs in vivo for a certain period of time. These may provide a novel choice for clinical treatment of peripheral nerve defects.

Glycoprotein-glycoprotein Receptor Binding Detection Using Bioluminescence Resonance Energy Transfer.

The glycoprotein receptors, members of the large G protein-coupled receptor family, are characterized by a large extracellular domains responsible for binding their glycoprotein hormones. Hormone-receptor interactions are traditionally analyzed by ligand-binding assays, most often using radiolabeling but also by thermal shift assays. Despite their high sensitivity, these assays require appropriate laboratory conditions and, often, purified plasma cell membranes, which do not provide information on receptor localization or activity because the assays typically focus on measuring binding only. Here, we apply bioluminescence resonance energy transfer in living cells to determine hormone-receptor interactions between a Gaussia luciferase (Gluc)-luteinizing hormone/chorionic gonadotropin receptor (LHCGR) fusion and its ligands (human chorionic gonadotropin or LH) fused to the enhanced green fluorescent protein. The Gluc-LHCGR, as well as other Gluc-G protein-coupled receptors such as the somatostatin and the C-X-C motif chemokine receptors, is expressed on the plasma membrane, where luminescence activity is equal to membrane receptor expression, and is fully functional. The chimeric enhanced green fluorescent protein-ligands are properly secreted from cells and able to bind and activate the wild-type LHCGR as well as the Gluc-LHCGR. Finally, bioluminescence resonance energy transfer was used to determine the interactions between clinically relevant mutations of the hormones and the LHCGR that show that this bioassay provides a fast and effective, safe, and cost-efficient tool to assist the molecular characterization of mutations in either the receptor or ligand and that it is compatible with downstream cellular assays to determine receptor activation/function.

A K homology (KH) domain protein identified by a forward genetic screen affects bundle sheath anatomy in Arabidopsis thaliana.

Because of their photosynthetic capacity, leaves function as solar panels providing the basis for the growth of the entire plant. Although the molecular mechanisms of leaf development have been well studied in model dicot and monocot species, a lot of information is still needed about the interplay of the genes that regulate cell division and differentiation and thereby affect the photosynthetic performance of the leaf. We were specifically interested in understanding the differentiation of mesophyll and bundle sheath cells in Arabidopsis thaliana and aimed to identify genes that are involved in determining bundle sheath anatomy. To this end, we established a forward genetic screen by using ethyl methanesulfonate (EMS) for mutagenizing a reporter line expressing a chloroplast-targeted green fluorescent protein (sGFP) under the control of a bundle sheath-specific promoter. Based on the GFP fluorescence phenotype, numerous mutants were produced, and by pursuing a mapping-by-sequencing approach, the genomic segments containing mutated candidate genes were identified. One of the lines with an enhanced GFP fluorescence phenotype (named ELEVATED BUNDLE SHEATH CELLS SIGNAL 1 [ebss1]) was selected for further study, and the responsible gene was verified by CRISPR/Cas9-based mutagenesis of candidate genes located in the mapped genomic segment. The verified gene, At2g25970, encodes a K homology (KH) domain-containing protein.

Evaluation of heterologous expression in Pichia pastoris of Pine Weevil TRPA1 by GFP and flow cytometry.

BACKGROUND: The wasabi receptor, also known as the Transient Receptor Potential Ankyrin 1 (TRPA1) ion channel, is a potential target for development of repellents for insects, like the pine weevil (Hylobius abietis) feeding on conifer seedlings and causing damage in forestry. Heterologous expression of TRPA1 from pine weevil in the yeast Pichia pastoris can potentially provide protein for structural and functional studies. Here we take advantage of the Green Fluorescent Protein (GFP) tag to examine the various steps of heterologous expression, to get more insight in clone selection, expression and isolation of the intact purified protein. RESULTS: The sequence of HaTRPA1 is reported and GFP-tagged constructs were made of the full-length protein and a truncated version (Δ1-708 HaTRPA1), lacking the N-terminal ankyrin repeat domain. Clones were screened on GFP expression plates, induced in small liquid cultures and in fed-batch cultures, and evaluated by flow cytometry and fluorescence microscopy. The screening on plates successfully identifies low-expression clones, but fails to predict the ranking of the best performing clones in small-scale liquid cultures. The two constructs differ in their cellular localization. Δ1-708 HaTRPA1 is found in a ring at the perimeter of cell, whereas HaTRPA1 is forming highly fluorescent speckles in interior regions of the cell. The pattern is consistent in different clones of the same construct and persists in fed-batch culture. The expression of Δ1-708 HaTRPA1 decreases the viability more than HaTRPA1, and in fed-batch culture it is clear that intact cells first express Δ1-708 HaTRPA1 and then become damaged. Purifications show that both constructs suffer from degradation of the expressed protein, but especially the HaTRPA1 construct. CONCLUSIONS: The GFP tag makes it possible to follow expression by flow cytometry and fluorescence microscopy. Analyses of localization, cell viability and expression show that the former two parameters are specific for each of the two evaluated constructs, whereas the relative expression of the constructs varies with the cultivation method. High expression is not all that matters, so taking damaged cells into account, something that may be linked to protein degradation, is important when picking the most suitable construct, clone, and expression scheme.

[Visualization of mitochondrial dynamics in tomato based on green fluorescent protein].

This study aimed to visualize the morphological features and dynamic changes of tomato mitochondria to provide a basis for the study of its mitochondrial functions. In this study, transgenic tomatoes expressing mitochondria-localized green fluorescent protein (mitochondria-GFP, Mt-GFP) were obtained by Agrobacterium-mediated genetic transformation. The color, hardness, soluble solids, acidity content, respiration rate, and ethylene production of the transgenic Mt-GFP tomato fruits were determined at the stage of mature green, breaker, and 3, 6, 9 days after breaker, while the wild-type tomato fruits were used as a control. As expected, Mt-GFP recombinant protein did not affect the ripening process, but induced the increased acidity of tomato fruits. The accumulations of Mt-GFP protein in tomato leaves and fruits were successfully verified by Western blotting. The morphological characteristics of mitochondria in flower, leaf and fruit cells as well as the dynamic changes of mitochondria in flower cells were clearly observed and studied under confocal laser microscope. The development of transgenic Mt-GFP tomato plants helps the visualization of tomato mitochondria and provides good research materials for the study of mitochondrial function during tomato development and fruit ripening.

Enhanced Detection of Estrogen-like Compounds by Genetically Engineered Yeast Sensor Strains.

The release of endocrine-disrupting compounds (EDCs) to the environment poses a health hazard to both humans and wildlife. EDCs can activate or inhibit endogenous endocrine functions by binding hormone receptors, leading to potentially adverse effects. Conventional analytical methods can detect EDCs at a high sensitivity and precision, but are blind to the biological activity of the detected compounds. To overcome this limitation, yeast-based bioassays have previously been developed as a pre-screening method, providing an effect-based overview of hormonal-disruptive activity within the sample prior to the application of analytical methods. These yeast biosensors express human endocrine-specific receptors, co-transfected with the relevant response element fused to the specific fluorescent protein reporter gene. We describe several molecular manipulations of the sensor/reporter circuit in a Saccharomyces cerevisiae bioreporter strain that have yielded an enhanced detection of estrogenic-like compounds. Improved responses were displayed both in liquid culture (96-well plate format) as well as in conjunction with sample separation using high-performance thin-layer chromatography (HPTLC). The latter approach allows for an assessment of the biological effect of individual sample components without the need for their chemical identification at the screening stage.

GFP Transfection Alters Protein Expression Patterns in Prostate Cancer Cells: A Proteomic Study.

PURPOSE: Green Fluorescent Protein is widely used as a cellular marker tool, but its potential influence on cells has been questioned. Although the potential off-target effects of GFP on tumor cells have been studied to some extent, the findings at the molecular level are insufficient to explain the effect of GFP expression on the tumorigenic capacity of cancer cells. Here, we aimed to investigate the effect of GFP expression on the tumorigenicity of PC3 prostate cancer cells. METHODS: Using GFP-expressing and wild-type PC-3 cells, xenograft models were generated in athymic BALB/C mice. To identify differentially expressed proteins, the change in cells proteome was investigated by label-free quantification with nano-high performance liquid chromatography to tandem mass spectrometry (nHPLC-MS/MS). Proteins that showed significantly altered expression levels were evaluated using the bioinformatics tools. RESULTS: Unlike the wild-type PC-3 cells, GFP-expressing cells failed to develop tumor. Comparative proteome analysis of GFP-expressing cells with WT PC-3 cells revealed a total of 216 differentially regulated proteins, of which 98 were upregulated and 117 were downregulated. CONCLUSION: Upon GFP expression, differential changes in several pathways including the immune system, translational machinery, energy metabolism, elements of cytoskeletal and VEGF signaling pathway were observed. Therefore, care should be taken into account to prevent reporting deceitful mechanisms generated from studies utilizing GFP.

Analysis of the Interaction Between Mucin and Green Fluorescent Protein (GFP)-Tagged Galectin-2 Using a 96-Well Plate.

Due to a significant proportion of glycans binding to the peptide (constituting approximately 50-90% of the molecular weight), analyzing the interaction between the entire mucin molecule and its recognition protein (lectin) can be challenging. To address this, we propose a semiquantitative approach for measuring the interaction between mucin and lectin, which involves immobilizing mucin in a 96-well plate and subsequently adding lectin tagged with green fluorescent protein.

Metabolome analysis of metabolic burden in Escherichia coli caused by overexpression of green fluorescent protein and delta-rhodopsin.

Overexpression of proteins by introducing a DNA vector is among the most important tools for the metabolic engineering of microorganisms such as Escherichia coli. Protein overexpression imposes a burden on metabolism because metabolic pathways must supply building blocks for protein and DNA synthesis. Different E. coli strains have distinct metabolic capacities. In this study, two proteins were overexpressed in four E. coli strains (MG1655(DE3), W3110(DE3), BL21star(DE3), and Rosetta(DE3)), and their effects on metabolic burden were investigated. Metabolomic analysis showed that E. coli strains overexpressing green fluorescent protein had decreased levels of several metabolites, with a positive correlation between the number of reduced metabolites and green fluorescent protein expression levels. Moreover, nucleic acid-related metabolites decreased, indicating a metabolic burden in the E. coli strains, and the growth rate and protein expression levels were improved by supplementation with the five nucleosides. In contrast, two strains overexpressing delta rhodopsin, a microbial membrane rhodopsin from Haloterrigena turkmenica, led to a metabolic burden and decrease in the amino acids Ala, Val, Leu, Ile, Thr, Phe, Asp, and Trp, which are the most frequent amino acids in the delta rhodopsin protein sequence. The metabolic burden caused by protein overexpression was influenced by the metabolic capacity of the host strains and the sequences of the overexpressed proteins. Detailed characterization of the effects of protein expression on the metabolic state of engineered cells using metabolomics will provide insights into improving the production of target compounds.

Covalent immobilization of human serum albumin on cellulose acetate membrane for scavenging amyloid beta - A stepping extracorporeal strategy for ameliorating Alzheimer's disease.

Alzheimer's disease (AD) is a neurodegenerative disorder characterized by interrupted neurocognitive functions and impaired mental development presumably caused by the accumulation of amyloid beta (Aß) in the form of plaques. Targeting Aß has been considered a promising approach for treating AD. In the current study, human serum albumin (HSA), a natural Aß binder, is covalently immobilized onto the surface of a cellulose acetate (CA) membrane to devise an extracorporeal Aß sequester. The immobilization of HSA at 3.06 ± 0.22 µg/mm2 of the CA membrane was found to be active functionally, as evidenced by the esterase-like activity converting p-nitrophenyl acetate into p-nitrophenol. The green fluorescent protein-Aß (GFP-Aß) fusion protein, recombinantly produced as a model ligand, exhibited characteristics of native Aß. These features include the propensity to form aggregates or fibrils and an affinity for HSA with a dissociation constant (KD) of 0.91 µM. The HSA on the CA membrane showed concentration-dependent sequestration of GFP-Aß in the 1-10-µM range. Moreover, it had a greater binding capacity than HSA immobilized on a commercial amine-binding plate. Results suggest that the covalent immobilization of HSA on the CA surface can be used as a potential platform for sequestering Aß to alleviate AD.

Optimization of medium components for protein production by Escherichia coli with a high-throughput pipeline that uses a deep neural network.

To optimize rapidly the medium for green fluorescent protein expression by Escherichia coli with an introduced plasmid, pRSET/emGFP, a single-cycle optimization pipeline was applied. The pipeline included a deep neural network (DNN) and mathematical optimization algorithms with simultaneous optimization of 18 medium components. To evaluate the DNN data sampling method, two methods, orthogonal array (OA) and Latin hypercube sampling (LHS), were used to design 64 initial media for each sampling method. The OA- and LHS-based data sampling resulted in green fluorescent protein fluorescence intensities of 0.088 × 103-1.85 × 104 and 3.30 × 103-1.50 × 104, respectively. Fifty DNN models were built using the OA and LHS datasets. Hold-out validation was performed using 15 % test of OA and LHS data. Mean square errors of the DNN models were 0.015-0.64, indicating the estimation accuracies were sufficient. However, the sensitivities of components in the DNN models varied and were grouped into six major classes by the index of k-means clustering. A representative model was selected for each class. Mathematical optimization algorithms using Bayesian optimization and genetic algorithm were applied to the representative models, and representative optimized medium (OM) compositions were selected by k-means clustering from the proposed OMs. A total of 54 OMs were obtained from the OA and LHS datasets. In the validating cultivation, the best OMs of OA and LHS were 2.12-fold and 2.13-fold higher, respectively, than those of the learning data.