Insights on the virulence of swine respiratory tract mycoplasmas through genome-scale metabolic modeling

Background The respiratory tract of swine is colonized by several bacteria among which are three Mycoplasma species: Mycoplasma flocculare, Mycoplasma hyopneumoniae and Mycoplasma hyorhinis. While colonization by M. flocculare is virtually asymptomatic, M. hyopneumoniae is the causative agent of enzootic pneumonia and M. hyorhinis is present in cases of pneumonia, polyserositis and arthritis. The genomic resemblance among these three Mycoplasma species combined with their different levels of pathogenicity is an indication that they have unknown mechanisms of virulence and differential expression, as for most mycoplasmas. Methods In this work, we performed whole-genome metabolic network reconstructions for these three mycoplasmas. Cultivation tests and metabolomic experiments through nuclear magnetic resonance spectroscopy (NMR) were also performed to acquire experimental data and further refine the models reconstructed in silico. Results Even though the refined models have similar metabolic capabilities, interesting differences include a wider range of carbohydrate uptake in M. hyorhinis, which in turn may also explain why this species is a widely contaminant in cell cultures. In addition, the myo-inositol catabolism is exclusive to M. hyopneumoniae and may be an important trait for virulence. However, the most important difference seems to be related to glycerol conversion to dihydroxyacetone-phosphate, which produces toxic hydrogen peroxide. This activity, missing only in M. flocculare, may be directly involved in cytotoxicity, as already described for two lung pathogenic mycoplasmas, namely Mycoplasma pneumoniae in human and Mycoplasma mycoides subsp. mycoides in ruminants. Metabolomic data suggest that even though these mycoplasmas are extremely similar in terms of genome and metabolism, distinct products and reaction rates may be the result of differential expression throughout the species. Conclusions We were able to infer from the reconstructed networks that the lack of pathogenicity of M. flocculare if compared to the highly pathogenic M. hyopneumoniae may be related to its incapacity to produce cytotoxic hydrogen peroxide. Moreover, the ability of M. hyorhinis to grow in diverse sites and even in different hosts may be a reflection of its enhanced and wider carbohydrate uptake. Altogether, the metabolic differences highlighted in silico and in vitro provide important insights to the different levels of pathogenicity observed in each of the studied species. Electronic supplementary material The online version of this article (doi:10.1186/s12864-016-2644-z) contains supplementary material, which is available to authorized users.

The average molecular weight of a nucleotide residue (inside DNA) was calculated from the molecular weight of nucleotides (Table ST3). Based on an average molecular weight of 306.82 g/mol of DNA residue, we concluded that 0.06g of DNA corresponded to 0.1955 mmol of DNA per gdW. The concentration of each base in mmol/gdW was estimated and may be found in Table 3.

RNA composition
The RNA composition was calculated similarly to DNA. The GC and AT content was based only on the open reading frames (ORFs) annotated in genbank files (Table ST4). An average nucleotide molecular weight inside the RNA molecules was calculated and can be found in Table ST5.  Based on an average molecular weight of 318.68 g/mol of RNA residue, we concluded that 0.12g of RNA, corresponded to 0.3765 mmol of RNA per gdW; the concentration of each base in mmol/gdW was estimated and is found in Table 3.

Amino acid composition
Amino acid composition accounted for all cellular proteins and was estimated by sequence analysis of translated mRNA (Table ST6). The average molecular weight of an amino acid residue was calculated and can be found in (Table ST7). ST6: Amino acid content from all analyzed strains. Based on an average molecular weight of 114.99 g/mol of amino acid residue, we concluded that 0.55g of proteins corresponded to 4.78 mmol of proteins per gdW; the concentration of each amino acid in mmol/gdW was estimated and may be found in Table 3.

Lipid composition
Lipids were divided in three categories: sterols, phospholipids and glycolipids [maniloff1992,Kornspam2012]. Phospholipids and glycolipids were separated into two fractions: elementary portion and fatty acid radicals ( Figure ST1). Based on the fatty acid composition [Chen1992,maniloff1992,Kornspam2012], we were able to estimate an average fatty acid molecular weight to incorporate into the elementary lipids portion. The fatty acids associated with elementary portions of lipids seem to mimic the total fatty acid composition of the organism and are dependent on the composition of the culture medium [Dahl1980, maniloff1992].  Table ST10. Multiple combinations of fatty acids and elementary portions can occur, this is why we treated each of them separately.
The main difference between M. hyorhinis and M. hyopneumoniae is that M. hyorhinis does not possess glycolipids [maniloff1992] while both glycolypids, monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG) were previously detected in M. hyopneumoniae [Chen1992]. Since no information was available for M. flocculare, we extrapolated the data from M. hyopneumoniae and included both glycolipids in its biomass composition.     (Table ST9) and the average fatty acid molecular weight in the correct quantities (as described in Figure ST1). ** When present, the percentage of phospholipids equals the combined percentages of neutral and glycolipids. When absent, neutral lipid content equals the phospholipid content Based on an average molecular weight of 682.0 g/mol for M. hyorhinis, we estimated that 0.15g of lipids corresponded to 0.220 mmol of lipids per g DW; the concentration of each elementary lipid and each fatty acid in mmol/g DW was estimated and may be found in Table 3. Based on an average molecular weight of 789.3 g/mol for M. hyopneumoniae and M. flocculare, we estimated that 0.15g of lipids corresponded to 0.19 mmol of lipids per g DW; the concentration of each elementary lipid and each fatty acid in mmol/g DW was estimated and may be found in Table 3.

Ions and Cofactors
We adapted the ions and cofactors composition from the metabolic network available for M. pneumoniae [Wodke2013]. The concentration of each compound can be found directly in Table 3.

Carbohydrate composition
The structural unit for carbohydrate/polysaccharide fraction was fixed only as glucose for simplicity reasons. However, the polysaccharide composition in these species was not yet determined. The concentration of glucose can be found directly in Table 3.

Biomass Reaction Assembly
In order to create the biomass, the cell must unwind and replicate DNA, transcribe and degrade RNA, translate and modify proteins, among others. The approximate costs for several maintenance functions associated with growth (GAM) were calculated and are presented in Table ST11. * To adjust the ATP requirements and accommodate tRNA charging directly into the models, activation and incorporation were discounted 2 ATPs per umol of protein.
We also have to take into consideration that for every polymerization step (and every ATP used), we have extra substrates and products. For instance, even if we consider that in the DNA, cytosine is in the form of dCMP, the reaction in vivo uses a dCTP as substrate and produces a pyrophosphate (PPi), incorporating a molecule of dCMP. n dNTPs + n ATP => DNA + n PPi n NTP + n ATP => RNA + n PPi n amino acids + n ATP => 1 protein + n ADP + n Pi + nH2O From all these estimated quantities, we were able to assemble the biomass reaction into the following form: Biomass Precursors + 11,53 ATP + 6,11 H2O + Charged tRNAs => 1 Biomass + 11,46 ADP + 11,45 Pi + 11,46 H+ + 0,57 PPi + Uncharged tRNAs where the stoichiometric coefficients for all biomass precursors can be found in Table 3 (with the exception of ATP, Pi and Ppi, which were adapted for ATP maintenance functions).