Characterisation and insight into the metabolism of oleaginous sub-Antarctic Corynebacteriales through genomic, lipidomic and proteomic analysis

A group of Actinobacteria, termed the mycolata within the order Corynebacteriales, are of industrial interest due to their potential to catabolise diverse organic and inorganic compounds (including toxins), tolerate high substrate concentrations in production-scale fermentation, and utilise waste materials to produce value-added products. The defining feature of this group is possession of a thick layer of long-chained lipids, the mycolic acids (MA), in their cell wall. MA are formed uniquely in the Corynebacteriales by the head-tohead condensation of fatty acids (FA), which occur in the cell wall as free or covalently bound to arabinogalactan. Together with other surface lipids, MA form a barrier which protects cells from noxious chemicals and enzymatic assault. Some members of this group are known to accumulate large amounts of triacylglycerols (TAG) as a storage compound, which makes them attractive targets for the manufacture of biodiesel. To find potential TAG producers which may have attractive biotechnological traits, six sub-Antarctic bacterial isolates from Macquarie Island were selected for further biochemical and genetic characterisation. These strains were previously screened for lysozyme resistance and were presumptive Actinobacteria. 16S rRNA gene sequencing determined that the strains belonged to two genera, Williamsia and Rhodococcus. Bacterial growth rates were determined at culture temperatures of 0 to 45 vÄv¿C and growth temperature ranges were predicted using a nonlinear model. All six isolates were psychrotolerant, with predicted Tmin and Tmax ranges of -12 to 37.5 vÄv¿C. Carbohydrate and nitrogen utilisation testing indicated that all strains exhibited a preference for D-fructose and better growth was observed with ammonium sulphate as N source, whilst four strains could utilise glycerol, a common industrial waste product and therefore an abundant carbon source suitable for industrial-scale fermentation. Carbon chain lengths were between 40-60 for MA and 30-60 for TAG analysed using ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS), and 12-23 for FA analysed using gas chromatography-mass spectrometry (GC-MS). The genomes of all six bacterial strains were sequenced using MiSeq Illumina technology after optimising culture conditions which facilitated the extraction of high molecular weight DNA. Whole genome sequences of all six strains were assembled using ABySS software and annotated using RAST and the NCBI Prokaryotic Genome Annotation Pipeline. Based on whole genome cluster analysis (pfam function), average nucleotide identity (ANI) and biochemical characteristics, the two Williamsia strains, 1135 and 1138, and two Rhodococcus strains, 1163 and 1168, were identified as novel species; another two Rhodococcus strains, 1139 and 1159, were identified as R. qingshengii and R. erythropolis, respectively. Genome sequences of R. qingshengii 1139 and Williamsia sp. 1138 were analysed manually to identify the genes involved in carbohydrate, nitrogen and MA transportation system. It was hypothesized that lack of glucose-specific phosphotransferase systems (PTS) in the genome of Williamsia sp. 1138 could be the reason for lower growth rate and biomass yield of the bacterial strain on D-glucose-supplemented medium. As some Rhodococcus species accumulate very high amounts of TAG, and they also have MA in the waxy cell envelop as well as membrane phospholipids (PL), it is important to understand how carbon is distributed across different lipid classes if attempting to optimise TAG biosynthesis under controlled fermentation conditions. Most analytical methods for identifying and quantifying different lipid classes involve different extraction, derivatisation, separation and detection techniques for each lipid class, so quantification of all rhodococcal lipids during TAG optimisation is normally not undertaken. Prior to lipid analysis on cells grown under different fermentation conditions, a single-step solvent system was developed using thin layer chromatography-flame ionisation detection (TLC-FID) to separate the major lipid classes in all six bacterial strains. Each lipid class was then quantified by calibrating with commercial standards, using surrogates for compounds if commercial standards were not available. A broad non-polar peak was observed as the major lipid class for all bacterial strains and the controls, Mycobacterium phlei and Corynebacterium glutamicum, which did not elute with the same retention time as the commercial free MA standard prepared from M. tuberculosis. The compounds in this peak were later identified using UPLC-MS/MS and nuclear magnetic resonance (NMR) analyses as MA bound to sugar fragments (likely arabinogalactan). The amount of carbohydrate in the non-polar lipid peak was extremely low (from NMR signal responses) so quantification of MA from peak area detected by TLC-FID was possible using a hydrocarbon surrogate as the standard. R. qingshengii 1139 was selected for proteomic and lipidomic analyses based on its rapid growth rate and higher biomass yields in shake-flask cultures, ability to utilise a broad range of carbon sources and synthesis of relatively high total lipids compared to other strains studied (total lipid of 29.8% of cell dry weight produced by strain 1139 compared to 22.6- 31.4% for the other strains, all of which grew more slowly with lower final biomass yields). Optimisation of carbon source concentration was undertaken in shake-flask cultures prior to culturing in controlled bioreactors using 4% D-fructose as carbon source and 2 g/L (NH\\(_4\\))2SO\\(_4\\) as N source. The experimental strategy was to determine the changes in the proteomes across the growth cycle as a preliminary study for future TAG optimisation. Results from TLC-FID analysis demonstrated that bound MA were the dominant lipid over the growth cycle with increasing concentrations per cell dry weight from mid-log to stationary phase but decreasing at late stationary phase. The concentration of polar lipids increased up to late log phase, then dropped at stationary and late stationary phases. In contrast, TAG concentration increased from mid-log to late stationary phase. Major changes were observed for the minor components of odd-chain-length FA analysed using GC-MS, however, the composition of MA and TAG analysed using UPLC-MS/MS did not change significantly over the growth cycle. A total of 1651 protein were detected from the tryptic digest of a cleared lysate of cells disrupted by bead beating (cell free extract) using nanoLC-MS/MS. A total 1297 out of 1651 proteins were obtained after filtering using Perseus software; two-sided t-test was performed for the proteins identified from late log, stationary and late stationary phase samples against mid-log phase samples. The proteins were grouped into 33 classes according to their function (based on BLAST/NCBI/RAST annotation) and a global heat map was constructed using the t-test values of proteins which demonstrated that lipid metabolism, amino acid degradation, nucleic acid degradation and detoxification systems were highly upregulated over the growth cycle. Detailed analysis of the highly upregulated protein groups indicated changes in relative abundance of proteins involved in the underlying pathways, which allowed predictions to be made about a suite of metabolic changes in cells as they moved from growth into stationary phase. These predictions suggest that amino acid and nucleic acid degradation supports bacterial survival at stationary phase, given that no residual D-fructose was detected by stationary phase, and also provides carbon flow into glycolysis to provide energy for further other metabolic pathways, including synthesis of TAG. It was also apparent that lipid degradation was occurring, which may support the hypothesis that cells recycled structural lipids for TAG production at the later stage of growth cycle under the assigned growth conditions. Furthermore, a nitrogen starvation condition was predicted in growth which induced the ureide pathway, which provides the NH3 through uric acid degradation also producesglyoxylate, as the majority of the enzymes of this pathway were detected only at stationary and late stationary phases. Proteins associated with the glyoxylate shunt were either detected or upregulated, suggesting that glyoxylate may be further metabolised through the glyoxylate shunt and other pathways to provide energy to cells. In addition to urea hydrolysing enzymes, a urea ABC transporter system was found highly upregulated with cell age indicating management of excess urea via cellular export. Analysis of proteins involved in defence systems revealed that this bacterial strain uses general stress proteins and glutathione-dependent glyoxalase systems to protect cells from oxidative stress, although chaperone proteins (DnaK and GroES/L) systems were not obviously upregulated at later stages of growth. This study helped us to understand how this oleaginous strain responds under nutritional stress conditions at the later stages of the growth cycle and how a shift of lipid classes between structural and storage lipids may be occurring in response to nutritional availability. The current study is not exhaustive, and additional information on the impacts of stationary phase transition can be gained from the current proteomic data sets; this is a matter for future further analysis, particularly relating to regulators detected in the data. Most of the previous studies demonstrated that TAG is the major lipid produced by Rhodococcus and related species under stressed conditions as quantified by fatty acid methyl esters analysis using GC-MS. However, our study demonstrated that bound MA were more abundant than other lipid classes under the culture conditions studied; this is the first study where all major lipid classes were quantified over the growth cycle. However, we could not separate...