Genome-wide high-throughput screening of interactive bacterial metabolite in the algal population using Escherichia coli K-12 Keio collection

Algae-bacteria interaction is one of the main factors underlying the formation of harmful algal blooms (HABs). The aim of this study was to develop a genome-wide high-throughput screening method to identify HAB-influenced specific interactive bacterial metabolites using a comprehensive collection of gene-disrupted E. coli K-12 mutants (Keio collection). The screening revealed that a total of 80 gene knockout mutants in E. coli K-12 resulted in an approximately 1.5-fold increase in algal growth relative to that in wild-type E. coli. Five bacterial genes (lpxL, lpxM, kdsC, kdsD, gmhB) involved in the lipopolysaccharide (LPS) (or lipooligosaccharide, LOS) biosynthesis were identified from the screen. Relatively lower levels of LPS were detected in these bacteria compared to that in the wild-type. Moreover, the concentration-dependent decrease in microalgal growth after synthetic LPS supplementation indicated that LPS inhibits algal growth. LPS supplementation increased the 2,7-dichlorodihydrofluorescein diacetate fluorescence, as well as the levels of lipid peroxidation-mediated malondialdehyde formation, in a concentration-dependent manner, indicating that oxidative stress can result from LPS supplementation. Furthermore, supplementation with LPS also remarkably reduced the growth of diverse bloom-forming dinoflagellates and green algae. Our findings indicate that the Keio collection-based high-throughput in vitro screening is an effective approach for the identification of interactive bacterial metabolites and related genes.


Results and discussion
High-throughput screening of interactive genes-mediating bacteria-algae interaction using Keio collection. As algae are an essential source of chemical energy for the ecosystem via oxygenic photosynthesis, they are considered a major primary producer of energy. However, eutrophication derived from high anthropogenic nutrient input and altered physical and biological interactions frequently induce the formation of HABs in aquatic environments 29,30 . Bacteria-algae interactions are one of the key mechanisms underlying the formation of algal blooms 31,32 . However, it is highly difficult and time-consuming to verify the bacteria-algae interactive metabolites, especially at the molecular level. To simplify the screening of interactive metabolites, we used the E. coli K-12 Keio collection. To verify the algal growth in response to the E. coli K-12 Keio collection, a green microalga C. vulgaris OW-01, which has shown optimal growth under lab-scale conditions, was used for screening. To specify the most relevant < 100 of algal growth-responsive genes, we have chosen the genes showing above 1.5-fold change than a control. Green microalga C. vulgaris OW-01 was effectively applied for our previous screening test using E. coli gene over-expression ASKA library 26 . We used the same method of our previous study to evaluate the bacteria-algae interaction using the E. coli K-12 Keio collection 26 . The axenic algal culture was confirmed by 18S rRNA sequence analysis and the bacterial colony formation test. To identify the test algal strain, the 18S rRNA was amplified, and the sequence obtained was compared against various sequences of other algal strains using BLAST and Mega software version 7.0 33 . As shown in Supplementary Fig. S1 (A)-based on the partial 18S rRNA-based phylogenetic tree-the isolated strain C. vulgaris OW-01 showed a close genetic relationship with C. vulgaris KMMCC FC-42 (accession no. HQ702285), Chlorella sp. ZB-2014 (accession no. KJ734869), Chlorella sp. YACCYB 497 (accession no. MH683919), C. sorokiniana (accession no. JX910111), and Chlorella lewinii (accession no. FM205861). Moreover, the microscopic image of the isolated strain showed a similar morphology with C. vulgaris 34 , confirming that the isolated strain was C. vulgaris OW-01 (Supplementary Fig. S1B). As shown in Supplementary Fig. S2A, the axenic condition of C. vulgaris was confirmed by electrophoresis of the universal 16S rDNA band. Whereas the rDNA bands were detected in xenic C. vulgaris and bacteria culture, the band was not detected in axenic C. vulgaris culture. Furthermore, no bacterial cells were detected by microscopic observation in the SYBR green stained samples. These results indicate axenic algal culture was maintained during the experiments.
After the strain was identified, we tested the E. coli Keio mutants using the C. vulgaris OW-01 strain to identify the bacteria-algae interactive genes and metabolites, as shown in Fig. 1. In our preliminary test, the kanamycin supplementation has not affected on algal growth response. Therefore, we supplemented the same concentration of kanamycin into algae-bacterial culture during the screening. www.nature.com/scientificreports/ After performing screening three times under optimized algal growth conditions, changes in chlorophyll a fluorescence were detected after 7 days of cultivation. By measuring the fluorescence of chlorophyll a, we identified bacterial genes of which the absence (deletion) resulted in a 1.5-fold increase in algal growth. The observation of white-yellow color in all screened experimental groups suggests increased population of bacterial cells. However, it is unclear what growth state these bacterial cells were in-due to the lack of available techniques for this purpose. This technical problem should be re-evaluated in further experiments. The Keio collection contains 3,985 gene knock-out mutants and a kanamycin resistance cassette that has been used to create a highly targeted single-gene disruption. Among the 3,985 genes, a total of 80 genes were selected via high-throughput screening, as shown in Supplementary Table S1. The function of each gene is also described in Supplementary Table S1.
The experimental groups showing enhanced microalgal growth (promoted relative fluorescence) were selected as potential genes coding for interactive algal growth inhibitors. To verify the related metabolites, we analyzed the relationships between the selected genes and the bacterial metabolite biosynthesis pathways. Fig. 2, we investigated the possible associations between related genes in the same pathways. Using the KEGG pathway database and EcoGene 3.0 software, we found that several genes, including lpxL, lpxM, kdsC, kdsD, and gmhB, were highly associated with LPS biosynthesis. LPS is produced by the amino sugar and nucleotide sugar metabolisms and the pentose phosphate pathway. These pathways synthesize LPS with the help of the detected genes. According to UniProt database (https ://www.unipr ot.org/), the lpxL and lpxX gene encodes lipid A biosynthesis lauroyltransferase and catalyzes the transfer of laurate from lauroyl-acyl carrier protein (ACP) to To verify the amount of LPS or lipooligosaccharide (LOS) synthesized by the selected mutants, the LPS content was analyzed as previously described (see "Materials and methods" section). As shown in Fig. 3A, all the selected bacterial mutants, including lpxL, lpxM, gmhB, kdsC, and kdsD, were found to generate significantly lower levels of LPS than the wild-type (about 8.66 μg mL −1 ). Our results indicate that the genes selected by high-throughput screening represent potentially relevant factors for the biosynthesis of bacterial LPS (or LOS). LPS is a well-known gram-negative bacterial endotoxin consisting of a hydrophobic domain lipid A, a core oligosaccharide (COS), and an O-polysaccharide (OPS) 35 . LPS has a molecular weight of 100,000 Da with a lipid A portion that is regarded as an in vivo and in vitro endotoxin. Furthermore, the polysaccharide portion of LPS provides immunogenicity, such that the molecular portions stimulate immune responses in cells 36 . The pathophysiological reactions of mammals to LPS, including fever, increased white blood cell counts, disseminated intravascular coagulation, hypotension, and inflammation, have been well-characterized in previous studies and reviews [36][37][38][39] . These biological activities stimulate both innate and adaptive immune responses in mammals via the production of cytokines and tumor necrosis factors (TNFs), as well as platelet-activating factors, phagocytosis, and immunoglobulins 36 .

Role of detected genes in lipopolysaccharide (LPS) biosynthesis pathway. As shown in
Similarly, LPS also induces immune responses in various terrestrial plants 40 . The immune response of Arabidopsis thaliana to OPS-COS isolated from the gram-negative bacterium Burkholderia cepacia was reported by Madala et al. 41 . Additionally, Bedini et al. 42 showed that bacterial LPS, comprised a l-rhamnose-rich backbone, and that synthetic oligorhamnans induce the expression of the pathogenesis-related-1 (PR-1) gene and suppress the hypersensitive response in a model of A. thaliana. As microalgae can also be classified as thallophytes, we hypothesized that LPS may induce growth and trigger an immune response similar to that observed in plants.
To evaluate the growth responses based on the LPS content, the microalga C. vulgaris OW-01 was exposed to different concentrations of LPS. As shown in Fig. 3B, algal cell growth was found to be decreased in a concentration-dependent manner when LPS was supplemented at concentrations > 10 µg mL −1 . These results indicate that bacterial LPS induces a toxic effect in microalgae. To verify the immune response in C. vulgaris, we analyzed the changes in the levels of reactive oxygen species (ROS) and the malondialdehyde (MDA) content of LPS-exposed algal cells. As shown in Fig. 4A, the ROS-dependent 2,7-dichlorodihydrofluorescein diacetate (DCFH-DA) fluorescence increased in a concentration-dependent manner when supplemented with LPS at a concentration Scientific RepoRtS | (2020) 10:10647 | https://doi.org/10.1038/s41598-020-67322-w www.nature.com/scientificreports/ ranging from 0 to 100 µg mL −1 . Similarly, the levels of the MDA, which is considered to be a marker for ROSinduced lipid peroxidation, was also enhanced in a concentration-dependent manner upon LPS supplementation, as shown in Fig. 4B. Although there is currently a lack of understanding regarding the role of LPS on the interaction between microalgae and bacteria, our results indicated that the algal growth-inhibitory effect of LPS was highly related to oxidative stress and cellular ROS-mediated lipid peroxidation.  Fig. 5, showing the different sizes and morphologies observed via light microscopy. Changes in the cell densities of the algal strains were detected after 7 days of cultivation supplemented with 0, 10, and 100 µg mL −1 LPS. As shown in Fig. 6, the majority of the algae tested showed decreased cell growth as a result of LPS supplementation. A concentration-dependent decrease in cell growth was also observed in the green microalgae C. sorokiniana and S. deserticola. Moreover, the harmful algal species A. tamarense and C. polykricoides were found to undergo a cell growth-inhibitory effect at an LPS concentration of 100 µg mL −1 . However, no significant growth response was observed in the cyanobacterium M. aeruginosa, as shown in Fig. 6F.

Effect of LPS supplementation on the growth of diverse algae.
Although further mechanistic studies are required, our results indicate that the algal growth response against LPS is species-specific. HABs caused by A. tamarense and C. polykrikoides are a well-recognized issue in the global aquaculture industry. The marine dinoflagellate A. tamarense is a paralytic shellfish poisoning toxin (PST) that produces harmful marine algae, whose blooms can have serious effects on ecosystems, the aquaculture industry, and public health 43,44 . C. polykrikoides is also considered major harmful algae. It is responsible for killing fish by producing reactive oxygen species (ROS), and its bloom has caused serious economic losses in the fishing industry in South  45 . To control the formation of HABs generated by these algal species, various biological methods using bacterial species have been developed in recent years [46][47][48][49][50][51][52] . To this end, the metabolite-related mechanistic study of the relationship between algae and bacteria can provide useful information for the development of an effective biological strategy for the control of HABs. However, the identification of bacterial interactive metabolites and their related genes is highly difficult and time-consuming. In this study, we performed high-throughput screening using the E. coli K-12 Keio mutant collection to demonstrate that LPS is a potential bacteria-microalgae interactive metabolite able to control microalgal growth responses.
Our findings indicate that this high-throughput method is a promising method for the effective verification of the bacterial interactive metabolites and related genes involved in the control of HAB formation, as well as algal biomass production.

conclusion
In the present work, we developed a high-throughput screening method for the identification of bacteriaalgae interactive metabolites using the E. coli K-12 Keio collection. The screening revealed that LPS serves as a possible algal growth-inhibiting bacterial interactive factor via oxidative stress. Furthermore, supplementation with LPS resulted in growth-inhibitory effects on other algal species, including C. sorokiniana, P. kessleri, S. deserticola, A. tamarense, and C. polykricoides. These results suggest that the E. coli K-12 Keio collection-based high-throughput screening can be an effective method for the verification of bacteria-microalgae interactions. However, more in-depth studies, investigating interactions of aquatic ecosystem-relevant bacterial species and bioactivities different types of LPS or LOS, along with the inclusion of each biosynthetic pathway for lipid-A, core and perhaps O-antigen on the algal growth are required to improve the bacteria-algae interaction mechanism under environmental conditions. High-throughput screening of algal growth response to bacterial metabolites using the Keio collection. The Keio collection is a set of single-gene knockout mutants of the E. coli K-12 strain BW25113.
This collection contains systematic and gene-deleted bacterial mutants. We compared a total of 3,985 E. coli single-cell knockouts lines with E. coli K-12 BW25113 serving as the wild-type to screen for the algal growth responses. Before the experiment, the Keio collection was placed in 96-well microplates (Eppendorf, Hamburg, Germany) and stored at − 80 °C after mixing with glycerol (final 40%, w/w). To evaluate the algal growth in response to the knock-out of specific genes in the Keio mutants, a 96-pin microplate replicator (Boekel Scientific, Feasterville PA, USA) was used for the co-culture of bacteria and algae, as shown in Fig. 1. Briefly, the stored glycerol stocks were allowed to thaw at room temperature for 30 min. These were then transferred into a 96-well microplate containing fresh medium and microalgae. For the co-culture of the bacteria and algae, modified algal/bacterial medium (BG11 with 5% Luria-Bertani (LB) media with 50 µg mL −1 kanamycin) was used, as proposed by Heo et al. 26 . The initial cell number of the E. coli mutants and microalgae was 10 5 CFU mL −1 and 5.0 × 10 6 cells mL −1 , respectively. The mixed culture was then stored under algal culture conditions (as described above) for 7 days. For the evaluation of algal growth, we measured the chlorophyll a-derived algal auto-fluorescence (excitation of 490 nm and emission 680 nm) using a fluorescence microplate reader (BioTek Instruments Inc., Winooski, VT, USA) after 7 days co-cultivation of bacteria and algae. The chlorophyll a readings were normalized by using the axenic control reading as zero [−50 < 0 (axenic control) < 50]. The gene knock-out clones that showed a 1.5-fold increase in relative fluorescence values compared to those of wild-type E. coli K-12 BW25113 cells were selected and matched with the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database and the EcoGene 3.0 database (https ://www.ecoge ne.org/) [56][57][58] . To prevent bacterial contamination, all the experimental procedures were performed triplicate using sterilized materials on a clean bench.

Determination of lipopolysaccharide (LPS) content.
To determine the LPS content, the E. coli mutants selected in the previous step were pre-cultured overnight in LB medium at 37 °C. Thereafter, the cultured cells (10 5 CFU mL −1 ) were incubated into 100 mL of fresh BG11 medium containing 5% (v/v) LB medium  www.nature.com/scientificreports/ cultured in F/2 medium 59,60 . All the algal strains were cultured in 500 mL of sterile algal medium in 1,000-mL bottle reactors for 4 days at an optimal temperature (freshwater algae at 25 °C, marine algae at 20 °C). To analyze the algal growth responses, 1 × 10 6 cells mL −1 of C. sorokiniana, 1 × 10 6 cells mL −1 of P. kessleri, 1 × 10 6 cells mL −1 of S. deserticola, 1 × 10 6 cells mL −1 of M. aeruginosa, 1 × 10 3 cells mL −1 of C. polykricoides, and 1.5 × 10 3 cells mL −1 of A. tamarense were treated with 0, 10, and 100 µg mL −1 of LPS in each optimal algal culture medium. After culturing, the algal cell density was analyzed using the direct cell method, as described above.
Detection of reactive oxygen species (ROS) based on 2,7-dichlorodihydrofluorescein diacetate (DCFH-DA) assay. To evaluate the levels of cellular ROS generated after LPS supplementation, LPStreated algal cells (10 mL) were harvested by centrifugation at 10,000×g for 1 min, followed by washing in PBS (pH 7.4) three times. Then, 10 μM of DCFH-DA was added to the algal cell suspension and incubated for 30 min in the dark at 25 °C. The DCFH-DA signals were then measured at an excitation wavelength of 485 nm and an emission wavelength of 535 nm, respectively 61 .
Determination of lipid peroxidation. The levels of cellular lipid peroxidation were determined using a thiobarbituric acid reactive substance (TBARS) assay to quantify the malondialdehyde (MDA) content. Briefly, 2 mL of algal culture was harvested by centrifugation at 10,000×g for 1 min. The resulting pellet was washed twice with PBS (pH 7.4), while the cell suspension was mixed with lysis buffer, comprising 1 μL of 10 mM butylated hydroxytoluene (BHT) solution and 99 μL of radio-immunoprecipitation assay (RIPA) buffer. After bead beating at 4,500×g for 1 min, 100 μL of 0.6 N trichloroacetic acid (TCA) solution was added to 100 μL of cell lysate before incubating at room temperature for 20 min. Centrifugation was then performed at 10,000×g for 5 min. The resulting supernatant was used for the TBARS assay. The TBARS assay was performed by mixing 150 μL of the sample with 75 μL of 0.5% (w/v) thiobarbituric acid (TBA) reagent. After incubating for 3 h at 50 °C, the absorbance of the culture was measured at 532 nm. The MDA content was measured based on the constructed standard curve of MDA 61 .
Statistical analysis. The analysis of variance (ANOVA) for experimental data sets was performed using JMP 4.0 software (SAS Institute Inc., Cary, NC). The significant difference was determined based on the magnitude of the F value (P < 0.01). When a significant F value was obtained for a treatment, separation of means was performed by determining Fisher's protected least significant difference (LSD) at P < 0.01. The analysis was conducted at least twice, with three replicates per well 26 . One-way ANOVA and subsequent t-test were performed to examine significant differences using SPSS 18.0 software (SPSS Inc., Chicago, IL, USA). All the experiments were tested in triplicate.