Hydrogen peroxide production and myo-inositol metabolism as important traits for virulence of Mycoplasma hyopneumoniae

Comparative genomics of glpO from glycerol metabolism

Highly conserved homolog genes to glpO from M. mycoides subsp. mycoides (EC 1.1.3.21) were found only in the genomes of M. hyopneumoniae and M. hyorhinis. Despite the annotation as a dehydrogenase in both M. hyopneumoniae and M. hyorhinis, we propose this enzyme to act as glycerol-3-phosphate oxidase (GlpO), using molecular oxygen as the final electron acceptor and producing DHAP and hydrogen peroxide. We therefore refer to the encoded protein in M. hyopneumoniae and M. hyorhinis as GlpO, rather than GlpD. The high similarity between these predicted proteins (Supporting Information Fig. S1A) may be an indication that this trait might be essential for the pathogenicity of these Mycoplasma species.

Particularly, the cytotoxicity of M. mycoides subsp. mycoides is considered to be related to the translocation of the hydrogen peroxide into the host cells (Bischof et al., 2008). This is presumably possible because of the close proximity to the host cells along with the integral membrane GlpO (Pilo et al., 2005, 2007). Different transmembrane prediction softwares (Hofmann, 1993; Combet et al., 2000; Krogh et al., 2001; Kahsay et al., 2005) identified putative transmembrane portions in the GlpO proteins from M. hyopneumoniae and M. hyorhinis (Supporting Information Fig. S1B). Similar results were reported for the homolog enzyme in M. mycoides subsp. mycoides (Pilo et al., 2005), and a recent proteomic study has detected GlpO from M. hyopneumoniae in surface-enriched extracts through LC-MS/MS (Personal communication from H. B. Ferreira, (Machado et al., 2018).

Pathogenic M. hyopneumoniae strains produce hydrogen peroxide from glycerol

Contrary to what we had anticipated, we were only able to detect the production of hydrogen peroxide from the two pathogenic strains of M. hyopneumoniae (7448 and 7422) in Friis medium, as can be seen in Fig. 1A. The attenuated strain from the same species (M. hyopneumoniae strain J), along with M. hyorhinis and M. flocculare did not produce detectable quantities of this toxic product. In order to verify if the amount of hydrogen peroxide produced was comparable between strains, we also counted the number of cells for each replicate. In this way, the two pathogenic strains produced approximately the same amount of hydrogen peroxide and had cell counts of the same order of magnitude (available in Supporting Information Table S1).

We also show (Fig. 1B) that the hydrogen peroxide produced by the M. hyopneumoniae strains 7448 and 7422 was dependent on the presence of glycerol in the incubation buffer.

Levels of glpO transcripts do not differ from pathogenic to attenuated strains of M. hyopneumoniae

We tested the three M. hyopneumoniae strains (7448, 7422 and J) in order to compare the mRNA expression levels of glpO gene by RT-qPCR. Since the transcript levels of normalizer genes were not comparable between strains, we used relative quantification normalized against unit mass; in our case, the initial amount of RNA. We chose one of the replicates from strain 7448 as the calibrator, and we were able to show (Fig. 2 and Supporting Information Table S2) that there was no significant difference in the transcript levels of glpO in all tested strains from M. hyopneumoniae.

Enzymes for the uptake and catabolism of myo-inositol are specific to M. hyopneumoniae strains

M. hyopneumoniae is the only reported species among the Mollicutes that contains genes involved in the catabolism of myo-inositol. Since Mycoplasma species seem to maintain a minimum set of essential metabolic capabilities, we decided to further investigate this pathway and the influence of its presence on the metabolism and pathogenicity of M. hyopneumoniae. The degradation of inositol can feed glycolysis with DHAP and also produces an acetyl coenzyme-A (AcCoA) (Fig. 3). A TU for the myo-inositol catabolism is present in all M. hyopneumoniae strains, with the exception of the gene that codes for the enzyme 6-phospho-5-dehydro-2-deoxy-d-gluconate aldolase (IolJ, EC 4.1.2.29), responsible for the turnover of 6-phospho-5-dehydro-2-deoxy-d-gluconate (DKGP) into malonate semialdehyde (MSA).

The gene encoding IolJ in other organisms is similar to the one coding for enzyme fructose-bisphosphate aldolase (Fba) from glycolysis (EC 4.1.2.13). There are two annotated copies of the gene fba in M. hyopneumoniae (fba and fba-1, Supporting Information Table S3). We performed homology and gene context analyses (with the use of the MGcV software (Overmars et al., 2013)), 3D comparative modelling and protein–ligand interaction analysis to check if either of them would be a suitable candidate for this activity.

The gene context and protein sequence alignment for 15 selected Fba homologs in Mollicutes can be seen in Supporting Information Figs S2 and S3. Comparative models for both copies of Fba from M. hyopneumoniae and previously characterized IolJ and Fba from Bacillus subtilis (Yoshida et al., 2008) were constructed based on available structures of Fba in PDB (Berman et al., 2000) (Fig. 4 and Supporting Information Table S4). Fba structures from Escherichia coli and Giardia intestinalis were used to gather more information about substrate binding (Supporting Information Fig. S3). The alignment shows a highly conserved zinc binding site (residues marked as ‘*’), essential for substrate binding and catalysis. Positions ‘a’, ‘b’, ‘c’, ‘d’ and ‘e’ surround the substrate cavity. The structural analysis suggests that the interaction mode of DKGP (substrate of IolJ) with the zinc ion of the active site is similar to that observed for FBP (fructose-1,6-bisphosphate, substrate of Fba). Nevertheless the substrate specificity is strongly dependent on the residues that form the substrate cavity.

While there seems to be several common features between Fba and IolJ (residues ‘c’, ‘d’, ‘e’ and ‘*’), residue ‘a’ appears to be essential for the substrate interaction with IolJ. This residue is generally occupied by an arginine (R52) in several putative IolJs from other organisms (Supporting Information Fig. S4), and absent in all predicted Fbas analysed in this study. From the predicted structures, the presence of this positively charged arginine in IolJ seems to disfavour the interaction with the phosphate group of FBP whilst it is complementary to the carboxyl group from DKGP.

In this way, the predicted structure of Fba-1 from M. hyopneumoniae resembles more the Fba structures from the experimentally solved Fbas in B. subtilis, E. coli and G. intestinalis. The annotated Fba from M. hyopneumoniae, on the other hand, seems to be more similar to the IolJ structure from B. subtilis. Although functional studies are needed to test this hypothesis, we propose that all enzymes needed for the myo-inositol catabolism are present in M. hyopneumoniae.

M. hyopneumoniae is able to uptake myo-inositol from the culture medium

In order to ascertain the ability of different bacteria to uptake myo-inositol, we used two different approaches. The first was the use of marked myo-inositol in complex medium and analysis by MS, and the second was to check the viability of cells (through ATP production) whenever myo-inositol was used as primary energy source.

When we tested if cells were able to uptake the marked myo-inositol, over the course of 48 h, we found no significant difference in M. flocculare and M. hyorhinis when compared to the control medium (CTRL), as observed in Fig. 5A. As expected, the concentrations of myo-inositol for both strains of M. hyopneumoniae after 48 h of growth were lower than the control medium. We also collected two extra time points for M. hyopneumoniae strain 7448 and CTRL: 8 h and 24 h of growth (Fig. 5B). In all time points, there is significant difference between the residual marked myo-inositol and the control medium, which implies that M. hyopneumoniae is able to uptake such carbohydrate from the medium. MS peak data is available in Supporting Information Table S5.

Since we had glucose and glycerol present in this complex medium analysed by MS, we also wanted to check if the viability of the different strains and species altered when myo-inositol was the primary energy source. For this, we incubated cells in myo-inositol defined medium (depleted with glucose and glycerol) for 8 hours and measured the amount of ATP these cells were able to produce. This in turn was directly related to the amount of viable cells after cultivation in the specific medium tested. Considering we do not know the energetic yield and efficiency of each strain and species, we could not directly compare the amount of ATP produced between different organisms. For this reason, growth in regular defined medium (with glucose) for each strain was used as a normalization control and the ratio of ATP production in both media was used to compare the viable cells between strains. Since there was no other energy source available in the medium and in accordance with our previous predictions and results, only M. hyopneumoniae cells remained viable (ranging from 75% to 280%) when compared to their control growth in the regular defined medium (Fig. 6 and Supporting Information Table S6). The viability of the other species in this medium was 11.5% for M. hyorhinis and 0.2% for M. flocculare. We also achieved similar results when comparing the growth in myo-inositol defined medium versus Friis medium (Supporting Information Fig. S5).

Glycerol metabolism and hydrogen peroxide production

Even though the GlpO enzyme was previously detected in proteomes from both pathogenic and attenuated strains of M. hyopneumoniae (232 and J) (Pinto et al., 2009; Pendarvis et al., 2014), only the pathogenic strains tested in our study (7448 and 7422) were able to produce detectable amounts of hydrogen peroxide in Friis medium (Fig. 1). To our knowledge, no other study up to now was able to show that M. hyopneumoniae strains were able to produce this toxic product in vitro (Maes et al., 2017). We also show here that the production of hydrogen peroxide in the pathogenic strains of M. hyopneumoniae is dependent on the presence of glycerol (Fig. 1B).

The metabolism of glycerol and the formation of hydrogen peroxide were described as essential for the cytotoxicity of lung pathogens M. mycoides subsp. mycoides (Vilei and Frey, 2001) and M. pneumoniae (Hames et al., 2009). Moreover, although both M. hyopneumoniae and M. flocculare can adhere to the cilia of tracheal epithelial cells in a similar way, only the adhesion of M. hyopneumoniae causes tissue damage (Young et al., 2000).

We showed that the difference in enzyme activity was not related to the expression levels of glpO gene from the strains tested (Fig. 2). We did not find any extreme differences in their aminoacid sequences either (Supporting Information Fig. S3). This could be an indication that either this enzyme undergoes post-translational modifications in order to be active and/or the availability of the substrate (glycerol) intracellularly might be a limiting step for its activity. Post-translational modifications have been extensively reported experimentally in several proteins of M. hyopneumoniae (Djordjevic et al., 2004; Burnett et al., 2006; Pinto et al., 2007; Seymour et al., 2010; Tacchi et al., 2016). From transcriptomic and proteomic literature data, we were not able to find any enlightening differences in this pathway between strains or species (Supporting Information Table S7).

As for the availability of intracellular glycerol, in our previous metabolic models, we predicted differences in the metabolism of glycerol among the three Mycoplasma species (Supporting Information Fig. S6). While M. hyopneumoniae has five different ways of uptaking glycerol (dehydrogenation of glyceraldehyde, ABC transport of glycerol and glycerol-phosphate, import of glycerophosphoglycerol and glycerophosphocholine), the other two species lack at least two reactions. This might also limit the rate of production of hydrogen peroxide in each species. In this way, the enhanced pathogenicity of M. hyopneumoniae over M. hyorhinis and M. flocculare may therefore also be due to hydrogen peroxide formation resulting from a higher uptake of glycerol as an energy source. Similarly, one reason that could partially explain why M. mycoides subsp. mycoides is highly pathogenic in comparison with the less pathogenic M. pneumoniae might be the greater intracellular availability of glycerol due to the presence of a specific and very efficient ABC transporter in M. mycoides subsp. mycoides (Hames et al., 2009).

Since the production of hydrogen peroxide was not reported as essential to the in vivo virulence of Mycoplasma gallisepticum (Szczepanek et al., 2014), more studies are needed to better understand the importance of this metabolism in M. hyopneumoniae. Moreover, future biochemical and functional studies are needed to prove that GlpO is indeed responsible for the activity proposed here and to check if the enzyme in attenuated strains/species is functional.

Myo-inositol uptake and catabolism

M. hyopneumoniae is the only Mycoplasma species with sequenced genome that has the genes for the catabolism of myo-inositol. Myo-inositol is an essential precursor for the production of inositol phosphates and inositol phospholipids in all eukaryotes (Gonzalez-Salgado et al., 2012). Myo-inositol is also widespread in the bloodstream of mammalians (Reynolds, 2009), which would make it a suitable energy source for bacteria in the extremely vascularized respiratory system. Previously, Mycoplasma iguanae was described to produce acid from inositol (Brown et al., 2006), but the methods used in that paper are not clear and there is no complete genome from this organism for us to draw any conclusions. Based on sequence homology, orthology, synteny and tridimensional analyses, we proposed a possible candidate for the missing enzyme IolJ in M. hyopneumoniae, namely a duplication of the fba gene from glycolysis. This functional divergence after duplication is particularly interesting in bacteria for which evolution was mostly driven by genome reduction. Another reported example of this event is the duplication of the trmFO gene in Mycoplasma capricolum and more recently in Mycoplasma bovis. The duplicated TrmFO in M. capricolum was reported to catalyze the methylation of 23S rRNA (Lartigue et al., 2014) while the duplicated copy in M. bovis has been described to act as a fibronectin-binding adhesin (Guo et al., 2017).

We showed here that M. hyopneumoniae was able to uptake marked myo-inositol from a complex culture medium (Fig. 5); in addition this was the only species that remained viable whenever myo-inositol was used as the primary energy source (Fig. 6). From our metabolic model predictions (Ferrarini et al., 2016), the use of myo-inositol would be much more costly than the uptake and metabolism of glucose, which corroborates the small uptake of myo-inositol in Friis medium (glucose-rich) (Fig. 5). This basal uptake of myo-inositol could also be an indication that this pathway is important not only for energetic yield. Supporting this idea, microarray studies on strain 232 showed that several genes (if not all) from the myo-inositol catabolism were differentially expressed during stress treatments: heat shock (downregulated) (Madsen et al., 2006), iron depletion (upregulated) (Madsen et al., 2006) and norepinephrine (downregulated) (Oneal et al., 2008). Moreover, a previous transcriptome profiling of M. hyopneumoniae (Siqueira et al., 2014) showed that all genes from the myo-inositol catabolism were transcribed under normal culture conditions. Furthermore, three genes from the pathway (iolB, iolC and iolA) belonged to the list of the 20 genes with the highest number of transcript reads. Besides the transcription of these genes, proteomic studies of M. hyopneumoniae strains 232 (Pendarvis et al., 2014), 7422, 7448 and J (Pinto et al., 2009; Reolon et al., 2014) (Supporting Information Table S7) showed that several enzymes from this pathway were present in normal culture conditions.

Indeed, myo-inositol has been extensively reported in several organisms as a signaling molecule (Downes and Macphee, 1990; Gillaspy, 2011). Moreover, the myo-inositol catabolism has been experimentally described as a key pathway for competitive host nodulation in the plant symbiont and nitrogen-fixing bacterium Sinorhizobium meliloti (Kohler et al., 2010). Host nodulation is a specific symbiotic event between a host plant and a bacterium. Kohler et al. (2010) showed that whenever inositol catabolism is disrupted (by single gene knockouts from the inositol operon), the mutants are outcompeted by the wild type for nodule occupancy. This means that genes for the catabolism of inositol are required for a successful competition in this particular symbiosis. Moreover, the authors were not able to find a suitable candidate for the IolJ activity. In our case, we proposed that the activity of the missing enzyme IolJ is taken over by a duplication of fba. We were able to find a similar duplication (also not inside the myo-inositol cluster) in the genome of S. meliloti 1021 (SM_b21192 and SM_b20199, both annotated as fructose-bisphosphate-aldolase, EC 4.1.2.13). This means that in at least one other symbiont that has the myo-inositol catabolism genes, there could exist a putative IolJ not close to the myo-inositol cluster, just as we proposed here.

Whether this entire pathway is functional in M. hyopneumoniae is yet to be tested and further experiments should take place to support this hypothesis. However, the ability of M. hyopneumoniae to persist longer in the swine lung if compared to the other two mycoplasmas might come from the fact that this species is able to uptake and process myo-inositol. Furthermore, the ability of M. hyopneumoniae to grow in diverse sites (Le Carrou et al., 2006) if compared to M. flocculare might also be due to this specific trait.

Concluding remarks

It is important to remember that even though M. hyopneumoniae is considered highly pathogenic, the three Mycoplasma species studied here are widespread in pig populations and can easily be found in healthy hosts (Fano et al., 2005; Pieters et al., 2010). However, the main question permeating this fact is: what causes the switch from a nonpathogenic Mycoplasma community to a pathogenic one? And what makes some strains pathogenic while others inflict no harm to the host cells?

Some strains of M. hyopneumoniae become less pathogenic in broth culture and, after serial passages, they lose their ability to produce gross pneumonia in pigs (Whittlestone, 2012). In a proteomic study comparing strains 232 and J, researchers have described that the attenuated strain J switches its focus to metabolism and therefore has developed better capabilities to profit from the rich culture medium while the ability to infect host cells becomes less important so that adhesion-related genes are downregulated (Li et al., 2009). This might be related to the fact that here we detected a higher production of ATP in this attenuated strain when compared to the pathogenic strains 7448 and 7422. Liu et al. (2013) have investigated genetic variations between M. hyopneumoniae strains 168 and attenuated 168-L and found out that almost all reported Mycoplasma adhesins were affected by mutations. Tajima and Yagihashi (1982) reported that capsular polysaccharides from M. hyopneumoniae play a key role in the interaction between pathogen and host. Indeed in several bacterial species it has been reported that the amount of capsular polysaccharide is a major factor in their virulence (Corbett and Roberts, 2009) and it decreases significantly with in vitro passages (Kasper et al., 1983). In this way, it is likely that the difference in pathogenicity between strains in M. hyopneumoniae does not solely depend on their metabolism, but also on their ability to adhere to the host.

A recent metagenomic analysis of community composition (Siqueira et al., 2017) has described that M. hyopneumoniae is by far the most prevalent species in both healthy and diseased hosts. The difficult isolation of Mycoplasma species from diseased lung extracts is due to the fact that, in culture, fast-growing bacteria will overcome the slow-growth of mycoplasmas (McKean et al., 1979). This means that, in vitro, the competition for an energy source between fast and slow-growing bacteria usually ends with an overpopulation of the fast growing ones. Given the fact that mycoplasmas survive for longer periods inside the host even in competition with other bacteria (Fano et al., 2005; Overesch and Kuhnert, 2017), we must assume that other factors exist and are usually not mimicked in cell culture.

While M. hyopneumoniae might cause no harm, depending mostly on the environment, the characteristics of the host and the composition of this dynamic lung microbiome, any unbalance in this system is probably capable of turning a nonpathogenic community into a pathogenic one. The final conclusion is that the disease is a multifactorial process depending on several elements that include intraspecies mechanisms, community composition, host susceptibility and environmental factors. One possibility is that the competition with fast-growing species could result in a lower carbohydrate concentration and that M. hyopneumoniae might have to overcome this environmental starvation with the uptake of glycerol or myo-inositol. Since the uptake of myo-inositol does not lead to the production of any toxic metabolite, it is more interesting for its persistence in the long run. Other bacteria will strongly compete for glucose and other related carbohydrates, while M. hyopneumoniae will have the entire supply of myo-inositol for itself. The uptake of glycerol as an energy source, on the other hand, will probably lead to the production of toxic hydrogen peroxide as reported in other Mycoplasma species. This toxic product combined with other toxins from the external bacteria in the system would most probably recruit immune system effectors. Since M. hyopneumoniae has efficient mechanisms of host evasion (Fano et al., 2005; Maes et al., 2017), the newly introduced and fast-growing bacteria might be eliminated faster and M. hyopneumoniae, in this way, would be able to persist longer than other species inside the host (as it is reported in vivo).

As mentioned before, virulence factors in Mycoplasma species cover a broader concept if compared to other species: they are genes not essential for in vitro conventional growth that are instead essential for optimal survival in vivo. From our M. hyopneumoniae metabolic models, neither the GlpO activity nor the uptake and metabolism of myo-inositol seem to be essential features for in vitro growth. However, we were able to show that they might be two metabolic traits important for the enhanced virulence of M. hyopneumoniae when compared to M. hyorhinis and M. flocculare and could be essential for its survival in vivo and directly affect its pathogenicity.

Mycoplasma cultivation

We used the following strains for experimental validation: M. hyopneumoniae strains 7448, 7422 (field isolates) and J (ATCC 25934), M. hyorhinis ATCC 17981 and M. flocculare ATCC 27716. Cells were cultivated in Friis medium (Friis, 1975) at 37°C for varying periods of time with gentle agitation in a roller drum.

Hydrogen peroxide detection

Hydrogen peroxide was detected in culture medium by the Amplex® Red Hydrogen Peroxide/Peroxidase Assay Kit (Invitrogen Cat. No A22188), according to the manufacturer's manual. M. hyopneumoniae, M. hyorhinis and M. flocculare were cultivated for 48 h in modified Friis medium (with no Phenol Red) and thereafter centrifuged. The supernatant was used for the hydrogen peroxide readings compared to a standard curve (Supporting Information Fig. S7). The medium without bacterial growth was used as negative control. We used biological and technical triplicates to infer the average amount of hydrogen peroxide produced, and the concentration was standardized based on the average number of cells from each culture. Statistical analyses were performed using GraphPad Prism 6 software by one-way ANOVA followed by Dunnett's multiple comparison test considering M. flocculare as a control (p < 0.05).

In order to determine if the hydrogen peroxide production was dependent on the glycerol metabolism, we used the Merckoquant Peroxide Test (Merck Cat. No 110011) with detection range of 0.5 to 25 μg of peroxide per ml of solution (as described in Hames et al. (2009). Fifteen ml of M. hyopneumoniae 7448 and 7422 strains were grown for 48 h in Friis medium, harvested by centrifugation at 3360 g and washed twice in the incubation buffer (67.7 mM HEPES pH 7.3, 140 mM NaCl, 7 mM MgCl₂). Cells were resuspended in 4 ml of incubation buffer and aliquots of 1 ml were incubated for 1 h at 37°C. To induce hydrogen peroxide production, either glycerol or glucose (final concentration 100 μM or 1 mM) was added to the cell suspension and samples were incubated at 37°C for additional 2 h. Hydrogen peroxide levels were measured using colorimetric strips according to the manufacturer's instructions. Aliquots without any added carbon source served as an incubation control. The statistical significance of the results was calculated using one-way ANOVA followed by Dunnett's multiple comparison test (p < 0.05). The results represent four biological replicates with at least two technical replicates each.

Mycoplasma cell count with flow cytometry

Mycoplasma cells cultivated for hydrogen peroxide detection were sedimented at 3360 g for 20 min at 4°C and washed three times with NaCl 0,9% (1x 3360 g for 20 min and 2x 3360 g for 4 min). Cells were resuspended in 1 ml of NaCl 0,9% and diluted 1:30 for flow cytometry readings in a Guava EasyCyte cytometer (Millipore, USA). Cells were characterized by side-angle scatter (SSC) and forward-angle scatter (FSC) in a four-decade logarithmic scale. Absolute cell counting was performed up to 5000 events and the samples were diluted until the cell concentration was below 500 cells/μl. The number of counts obtained was then converted to cells/ml.

Transcript levels of glpO with the use of real-time quantitative RT-PCR

Total RNA was isolated from 20 ml culture of M. hyopneumoniae strains 7448, 7422 and J grown at 37°C for 24 h. Cells were harvested by centrifugation at 3360 g for 15 min, resuspended in 1 ml of TRizol (Invitrogen, USA) and processed according to the manufacturer's instructions followed by DNA digestion with 50 U of DNaseI (Fermentas, USA). Absence of DNA in the RNA preparations was monitored by PCR assays. The extracted RNA was analysed by gel electrophoresis and quantified with the Qubit^TM system (Invitrogen, USA). A first-strand cDNA synthesis reaction was conducted by adding 500 ng of total RNA to 500 ng of pd(N)6 random hexamer (Promega, USA) and 10 mM deoxynucleotide triphosphates. The mixture was heated for 65°C for 5 min and then incubated on ice for 5 min. First-strand buffer (Invitrogen, USA), 0.1 M dithiothreitol and 200 U M-MLV RT (Moloney Murine Leukemia Virus Reverse Transcriptase – Invitrogen, USA) were then added to a total volume of 20 μl. The reaction was incubated at 25°C for 10 min and at 37°C for 50 min followed by 15 min at 70°C for enzyme inactivation. A negative control was prepared in parallel, differing only by the absence of the RT enzyme. Quantitative PCR (qPCR) assay was performed using 1:2.5 cDNA as template and Platinum SYBR Green qPCR SuperMix-UDG with ROX (Invitrogen, USA) with specific primers for glpO (5′GGTCGGGAACCTGCTAAAGC3′ and 5′CCAGACGGAAACATCTTAGTTGG3′) on StepOne Real-Time PCR Systems (Applied Biosystems, USA). The qPCR reactions were carried out at 90°C for 2 min and 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. A melting curve analysis was done to verify the specificity of the synthesized products and the absence of primer dimers. The amplification efficiency was calculated with the LinRegPCR software application (Ruijter et al., 2009). A relative quantification normalized against unit mass (500 ng of total RNA) was used to analyse the expression data with the equation: , where (Calviño et al., 2008) and MHP_7448 (Replicate 2) was chosen as calibrator. Statistical analyses were performed using GraphPad Prism 6 software by one-way ANOVA followed by Tukey's multiple comparison test (p < 0.05).

Comparative modelling and protein–ligand interaction analysis of Fba and IolJ

The SWISS-MODEL server (Schwede et al., 2003; Biasini et al., 2014) was used for template search and the comparative modelling for all Fba and IolJ proteins in this study. The best homology models were selected according to coverage, sequence identity, Global Model Quality Estimation (GMQE) and QMEAN statistical parameters (Benkert et al., 2009; Benkert et al., 2011). The Fba from M. hyopneumoniae along with IolJ and Fba from B. subtilis were modeled using the crystal structure of fructose 1,6-bisphosphate aldolase from Bacillus anthracis in complex with 1,3-dihydroxyacetonephosphate (PDB 3Q94) while Fba-1 from M. hyopneumoniae was modeled using the fructose-1,6-bisphosphate aldolase from Helicobacter pylori in complex with phosphoglycolohydroxamic acid (PDB 3C52). Both selected templates have the same resolution range (2.30 Å). Fba structures experimentally solved from E. coli (Hall et al., 1999) and G. intestinalis (Galkin et al., 2009) were used to include information about substrate binding in the active site. The DKGP and FBP ligands were drawn in the Avogadro version 1.1.1 (Hanwell et al., 2012) by editing the tagatose-1,6-biphosphate (TBP) molecule complexed with the Fba structure of G. intestinalis (PDB 3GAY). Each model was submitted to 500 steps of an energy minimization protocol using the universal force field (UFF). The DKGP and FBP molecules were inserted into the substrate binding sites of the acquisition models obtained by superposition of the models with the Fba structure of G. intestinalis.

Detection of marked myo-inositol through mass spectrometry

Determination of cell viability of M. hyopneumoniae in myo-inositol defined medium

All available strains were grown in Friis medium at 37°C for 48 h, sedimented by centrifugation at 3360 g for 20 min at 4°C, washed twice with ice cold PBS and inoculated in glucose regular defined medium (described in Ferrarini et al. (2016), supplemented with 5 g/l of succinate) or myo-inositol defined medium (regular defined medium depleted with glucose and glycerol and supplemented with 0.5 g/l of myo-inositol and 5 g/l of succinate). Viability of cells was measured by ATP production with live cells recovered after 8 h of growth in either media with a BacTiter-Glo^TM Microbial Cell Viability Assay Kit (Promega, USA) according to the manufacturer's manual. Luminescence was recorded in a SpectraMax MiniMax 300 Imaging Cytometer (Molecular Devices, USA) with an integration time of 0.5 s in an opaque-walled multiwell plate. Average ATP production was calculated with biological duplicates and technical triplicates. The ATP production of each strain was compared between regular defined medium and myo-inositol defined medium to determine the ratio of viable cells and to allow a comparison between strains. A 10-fold serial dilution of ATP was used as a standard curve (Supporting Information Fig. S8). Statistical analyses were performed using GraphPad Prism 6 software by one-way ANOVA followed by Tukey's multiple comparison test (p < 0.05).