Metabolomics Deciphered Metabolic Reprogramming Required for Biofilm Formation

Biofilm formation plays a key role in many bacteria causing infections, which mostly accounts for high-frequency infectious recurrence and antibiotics resistance. In this study, we sought to compare modified metabolism of biofilm and planktonic populations with UTI89, a predominant agent of urinary tract infection, by combining mass spectrometry based untargeted and targeted metabolomics methods, as well as cytological visualization, which enable us to identify the driven metabolites and associated metabolic pathways underlying biofilm formation. Surprisingly, our finding revealed distinct differences in both phenotypic morphology and metabolism between two patterns. Furthermore, we identified and characterized 38 differential metabolites and associated three metabolic pathways involving glycerolipid metabolism, amino acid metabolism and carbohydrate metabolism that were altered mostly during biofilm formation. This discovery in metabolic phenotyping permitted biofilm formation shall provide us a novel insight into the dissociation of biofilm, which enable to develop novel biofilm based treatments against pathogen causing infections, with lower antibiotic resistance.

Metabolite extraction. Biofilm samples were isolated from 50 ml of culture solution, and the supernatants in the culture medium were spun down at 3500 rpm for 10 minutes to harvest the planktonic cells. Firstly, each biofilm was clean with PBS for 3 times, and added 2 ml the ice-cold methanol to mix by homogenizer for 2 minutes, all procedures were repeated for 3 times, but the samples were frozen and thawed within liquid nitrogen during the process. Secondly, the samples were centrifuged under12000 rpm at 4° for 10 minutes. The supernatants were mixed with 800 ul of ice-cold acetonitrile for 15 minutes before lyophilization. The dried samples were stored in −80° for GC-MS and LC-MS assay.
Sample derivation for GC/MS based metabolome assay. The lyophilized samples were resuspended in 40 ul of methoxyamine hydrochloride in pyridine (20 mg/ml) and completely mixed for 90 minutes at 30 °C. N-Methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA) (90 ul) was added to each sample, then mixed further for 30 minutes at 37 °C, and they were undergoing vortex before incubation. The derivative samples were centrifuged at 12000 rpm for 10 minutes and then transferred into a 2 ml GC vial and tightly capped for metabolome assay.
Data acquisition and analysis of GC/MS based metabolome assay. Metabolome assay was carried out on an Agilent 7890-5975 C GC/MS system using helium as the carrier gas at a constant flow. A total of 1 ul of sample was injected into a 50 m × 250 μm × 0.25 μm DB5-MS column using split less injection. The injector temperature was at 275 °C. Front-Pressure, average speed, flow rate was set at 22.468 psi, 34.908 cm/sec and 1.5 mL/ minutes, respectively. The GC/MS data was firstly transformed to a digital data using XCMS online database and www.nature.com/scientificreports www.nature.com/scientificreports/ then identified the metabolites by combining AMDIS software and associated database, which based on their retention index (RI) values along with the corresponding mass spectra. The identified metabolites were listed in a data-matrix file that was incorporated with metabolite ID, sample ID and the abundance of each metabolites, it was used for further pattern-recognition and statistical analyses.
Data acquisition and analysis of LC/MS based metabolome assay. Lyophilized samples were resuspended in 100 ul water and was injected to a UFLC-MS/MS system (AB SCIEX API 3200 TQ MS System; UFLC-SHIMADZU, 20AD) equipped with an ESI source in both negative and positive ion-modes with an electrospray ionization voltage of 5500 V for positive mode and 4500 V for negative mode, nebulizer gas (air), turbo gas (air), curtain gas (nitrogen) setting at 50, 50 and 25 psi respectively. Chromatographic separation was carried out on a Waters XSelect HSS T3 column (2.1 * 100 mm, 3.5 um) with a gradient program as follows: 5-95% Acetonitrile (Mobile Phase-B) for 50 minutes at a flow rate of 0.4 ml/minute. The column temperature was maintained at 40 °C, Mobile-Phase-A: 0.1% formic acid in water and Mobile Phase-B: 0.1% formic acid in acetonitrile. AB SCIEX Analyst software was used for data acquisition and preprocessing. The targeted MRM method was developed based on the acquired reference compounds, which assisted in targeted profiling the known compounds of interest from the known metabolic pathways.
Statistical analysis. The data matrix was transferred to MetaboAnalyst 3.0 version online to engage in multivariate statistical analysis including unsupervised principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA) and heat-map overview. The loading plot (VIP plot) was used to discover and identify the differential metabolites, as well as the score plot was adopted to facilitate group classification. This brief method was referred to the publication 3 .

Results and Discussion
Morphological observation verified successful development of biofilm model. Since notable morphological shifts of biofilm, pellicle and planktonic population were revealed in previous study 1 . We were first to confirm whether biofilm pellicle model was successful or not in our study (Fig. 1B), we employed the cytological method including Crystal Violet Staining, TEM (transmission electron microscope) and SEM (scanning electron microscope), to investigate the intricate-structure phenotype of biofilm formation, as it was expectedly found that pellicle biofilm have an organized and complex extracellular matrix-structure that encased individual, and spatially segregated bacterial-subpopulations (Fig. 1E,F), but such phenotype disappeared in the relevant planktonic-subpopulation (Fig. 1C,D). This result was completely consistent with previous study 1 . The imaging assay surely confirmed the biofilm pellicle model was successfully established in this study. In addition, staining technologies such as crystal violet and congo red were used to quantitatively assess the biofilm formation that was also significantly different from the planktonic population (Fig. S1). Those phenotypic observations may suggest us there is modified metabolism in two conditions. Distinctly metabolic modifications were observed in both biofilm formation and associated planktonic-population. Metabolomics is an systems-biology driven omics method that was designed to global profiling all the small-molecule metabolites, whose level changes are capable of capturing and snapshotting different biological events and processes in cells with a diverse biological contexts 6,7 . To investigate reprogrammed-metabolism underlying biofilm formation, we explored one combinational strategy with untargeted and targeted metabolomics for deciphering the distinct metabolism that was critical for biofilm formation. Metabolome assay revealed that distinctly metabolic reprogramming during biofilm formation, which is significantly different from that in planktonic population (Fig. 2). From the score plot resulted from PLS-DA analysis of metabolome data harvested by LC/MS/MS system ( Fig. 2A), we can see principal component 1 (PC-1) accounted for 31.3% of the variance, with PC-2 explaining 34%. The score plot by GC/MS based metabolome assay ( Fig. 2B) indicated, principal component 1 (PC1) explained 58.9% of variance, with PC2 of 18.4%, which exhibit the greatest difference predominantly along the PC-1 analysis. Global metabolome result was also illustrated in Fig. 2C. Those results revealed remarkable metabolic-differentiation between biofilm formation and planktonic-subpopulation, while bacterial cells were switched to form biofilm, they could secret the extracellular polymeric substance (EPS), which can be used to protect biofilm to defense environment insults and difficult to dissolution. Although our TEM images only recognize the denser layer of fibers 1 (highlighted in red in Fig. 1E,F) surrounding bacteria in biofilm pellicle mode, many other components (eDNA, proteins and polysaccharides, etc) present in EPS structure 5-7 which were disappeared in planktonic population. We deduced that microbes are capable of sensing the density of microenvironment as well as releasing the signal to change the gene expression and further yielding the influence upon bacterial metabolism. Consequently, reprogrammed-metabolism would have feedback to the modification with morphological features of the biofilm. In addition, the data also suggested that there was a mechanistic association between biofilm metabolism and its' morphological feature.
Metabolic reprogramming triggered biofilm formation. Metabolic reprogramming is supposed to be the key driven-factor for biofilm formation even though such organism was structured by the involvements of multiple-layers molecules. We firstly demonstrated obvious modifications in both of small-molecule metabolism and morphological feature between biofilm and planktonic cells, however the profoundly metabolic mechanism remains largely unexplored. Here, we further combined the open-source databases with local database to identify the pivotal metabolites and associated metabolic pathways that have capacity to drive the biofilm formation while compared to planktonic cells. Unexpectedly, we successfully identified 38 differential metabolites, including amino acids, carbohydrates, organic acids, glycerol-derived, polyamine and uridine between biofilm pellicle and planktonic cell (Table S1). It was an amazing discovery as glycerol-derived metabolites (2-Phosphoglyceric acid, Glycerol 3-phosphate and D-Glyceraldehyde 3-phosphate) were significantly up-regulated during biofilm (2019) 9:13160 | https://doi.org/10.1038/s41598-019-49603-1 www.nature.com/scientificreports www.nature.com/scientificreports/ formation while the glycerol was decreased considerably. Similarly, we also found that the levels of carbohydrates were enhanced clearly except that the levels of D-Maltose and D-Glucose were decreased in biofilm formation in relevant to that in planktonic cells. The level changes of other differential metabolites in both modes were showed in Table S1 and Fig. 3. Next, to home those differential metabolites to their metabolic pathways, we subjected them to metaboanalyst online to annotate associated metabolic pathways (Fig. 4). What metabolism we expected that were mostly reprogrammed that involved glycolipid metabolism, amino acid metabolism and carbohydrate metabolism, which were observed to markedly drive biofilm formation as most of differential metabolites were up-regulated accordingly (highlighted in blue in Fig. 4).
It was reported that proteins and carbohydrates were the main components of biofilm formation 2 . Several amino acids were documented to account for biofilm detachment and maturation [8][9][10][11] . In addition, the mixture of D-Amino acids would lead to the biofilm desperation with Bacillus subtilis and further break down the biofilm by YqxM protein which regulated the amyloid fibers anchoring or off-anchoring to the host cells 10 . However, our results only found the levels of many L-amino acids rather than D-amino acids were sharply changed (Fig. 3C), which might reveal the L-amino acids may be also participated in biofilm formation induced by UTI89 11 while D-amino acids were catalyzed as dispensable nutrition for bacteria growth 12,13 . In addition, we also noticed that the level changes of different amino acids were extremely distinct between biofilm and planktonic cells. This unregularly phenomenon supported again that amino acids may specially play their individual roles in biofilm formation 14 . Metabolic reprogramming of amino acids in this study was confirmed again to mostly trigger biofilm formation.
AI-2 is a typically quorum-sensing (QS) molecule, it can regulate physiological process of biofilm formation as the density of bacterial cells changed 6 . The lsr protein is essential for importing and processing AI-2, which was noticed to be inhibited by high concentration of Glycerol-3-phosphate (G-3-P) 15 . We speculated that biofilm formation preferred to consume glycerol to biosynthesize G-3-P as it could regulate the expression of AI-2 to avoid biofilm dispersal. This was agreeable with our result that the level of glycerol during biofilm formation was decreased remarkably while the production of G-3-P was enhanced substantially (Fig. 3B). Moreover, the most carbohydrates were incorporated into FimH adhesion (mannose) to increase the adhesive capability of biofilm 16 or participate in the EPS formation. So bacteria would seek other nutrients to maintain their survival. The G-3-P was the key precursor of glyceraldehyde-3-phosphate, it was a key factor that connected the carbohydrate metabolism to glycerolipid metabolism, as well as was a candidate for providing the microbes with the necessary energy. Thus, our finding convinced that more G-3-P has been synthesized, via consuming more glycerol to maintain the survival of bacterial cells during biofilm formation.
Interestingly, our data showed that putrescine was expressed differentially between two modes of UTI89. This result was well supported by previous studies that putrescine plays a crucial role in modulating the biofilm formation triggered by some bacteria 7,17-19 . Another polyamine spermidine was observed to have a remarkable shift in   www.nature.com/scientificreports www.nature.com/scientificreports/ biofilm formation, which was not expected since spermidine was verified only to inhibit the growth of E. coli at PH 7, but not link to the biofilm formation 20 . In near future, we shall figure out to study further the mechanisms of this metabolite modulating biofilm formation.
Bacterial cells in biofilm mode have extraordinary survival ability 21 to resist the environmental stress largely via the protective EPS. Amino acid metabolism, carbohydras metabolism and glycolipid metabolism in bacteria were not only responsible for the biosynthesis and storage of accessible energy but also were utilized for synthesizing the secreted components such as amino acids, sugars, lipids, uridines and organic acids, which are essential for EPS production during biofilm formation 22 . Those metabolic pathways were highlighted in this study to have significantly metabolic reprogramming that triggered the biofilm formation from metabolic perspective (Fig. 4). Those findings indicated that biofilm formation is ought to yield a protective feature, by funneling metabolites into EPS synthesis, conveying the stronger protection against stress conditions via dominating the metabolic reprogramming. Biofilm network is supposed to build a anaerobic microenvironment for bacteria embedded in biofilm pellicle 23 , where bacterial cells were structured in a stable space. Thus, bacterial biofilm could secure more resources in order to survive well while compared to planktonic cells.

conclusion
In summary, this study was first to combine metabolomics with cytology methods together to better understand the biofilm formation in UTI89 strain.
Our data has revealed that obviously metabolic reprogramming triggered biofilm formation while compared to planktonic population as we found lots of small-molecule metabolites are critically essential for biofilm formation and dissociation induced by UTI89. Those differential metabolites and the associated metabolic pathways can be regarded as novel targets for the development of biofilm based treatments and antibiotic discovery against the UTI89 causing infection. More importantly, such effort could provide a novel insight into better understanding of the biofilm formation caused by different bacterial strains at metabolic level.