Intraoral Microbial Metabolism and Association with Host Taste Perception

Metabolomics has been identified as a means of functionally assessing the net biological activity of a particular microbial community. Considering the oral microbiome, such an approach remains largely underused. While the current knowledge of the oral microbiome is constantly expanding, there are several deficits in knowledge particularly relating to their interactions with their host. This work uses nuclear magnetic resonance spectroscopy to investigate metabolic differences between oral microbial metabolism of endogenous (i.e., salivary protein) and exogenous (i.e., dietary carbohydrates) substrates. It also investigated whether microbial generation of different metabolites may be associated with host taste perception. This work found that in the absence of exogenous substrate, oral bacteria readily catabolize salivary protein and generate metabolic profiles similar to those seen in vivo. Important metabolites such as acetate, butyrate, and propionate are generated at relatively high concentrations. Higher concentrations of metabolites were generated by tongue biofilm compared to planktonic salivary bacteria. Thus, as has been postulated, metabolite production in proximity to taste receptors could reach relatively high concentrations. In the presence of 0.25 M exogenous sucrose, increased catabolism was observed with increased concentrations of a range of metabolites relating to glycolysis (lactate, pyruvate, succinate). Additional pyruvate-derived molecules such as acetoin and alanine were also increased. Furthermore, there was evidence that individual taste sensitivity to sucrose was related to differences in the metabolic fate of sucrose in the mouth. High-sensitivity perceivers appeared more inclined toward continual citric acid cycle activity postsucrose, whereas low-sensitivity perceivers had a more efficient conversion of pyruvate to lactate. This work collectively indicates that the oral microbiome exists in a complex balance with the host, with fluctuating metabolic activity depending on nutrient availability. There is preliminary evidence of an association between host behavior (sweet taste perception) and oral catabolism of sugar.


Bacterial load
Final bacterial load was assessed post-incubation. Samples were vortexed to homogenise the bacterial content and sample was serially diluted ten-fold to 1:10 5 . Samples diluted to 1:10 3 and 1:10 5 were plated (20 µl) onto fastidious anaerobe agar with 5% defibrinated horse blood. Plates were incubated under anaerobic conditions for 48 hours, colonies counted and CFU/ml calculated.

Protein quantification
Bacterial cells were removed by centrifugation at 15,000 g for ten minutes at 4 °C. Samples were analysed by SDS-PAGE as previously described (Gardner and Carpenter 2019). Briefly, 12 µl buffered sample was added per lane, electrophoresed and stained with Coomassie Brilliant Blue R250 (Sigma, Gillingham, UK). Samples of the unincubated parotid saliva and incubated, PBS-inoculated parotid saliva were run on every gel. Destained gels were imaged with a ChemiDoc MP system (Biorad, Watford, UK) and analysed in ImageLab 5.2.1 (Bio-Rad Laboratories, Hercules, CA, USA). Total lane density of sample lanes and PBS-inoculated lanes relative to the unincubated parotid saliva lanes were measured. Bacterial pellets were retained for analysis confirming the protein changes observed in saliva (Supplementary Figure 2).

H-NMR spectroscopy
Centrifuged samples were prepared and analysed using internal standard as described (Gardner et al. 2018). All reagents and consumables were purchased from sigma. NMR buffer was prepared with 0.5 mM trimethylsilyl-[2,2,3,3,-2H 4 ]-propionate (TSP) standard, 0.2 M Na2HPO4 and 44 mM NaH2PO4 in 50% deuterium oxide (D2O) by volume. Sample (440 µl), centrifuged as for the protein analysis, was mixed with NMR buffer (110 µl) in 5 mm external diameter NMR tubes to give a final concentration of 0.1 mM TSP and 10% by volume D2O. Using a 600 MHz spectrometer (Bruker, Karlsruhe, Germany), operating at a proton frequency of 600.2 MHz, spectra were acquired at 25 °C using a Carr-Purcell-Meiboom-Gill (CPMG) spin-echo pulse sequence with presaturation to supress macromolecule resonances from the spectra. The total echo time was 64 ms with relaxation delay of 4 s and acquisition time of 2.32 s. Following four dummy scans, 128 transients were collected with 64,000 data points and spectral width of 20 ppm (−5 to 15 ppm). Spectra were automatically phased and baseline corrected with further manual adjustment where required. The control samples described for the protein analysis were also analysed by 1 H-NMR spectroscopy. An additional nonincubated sample of parotid saliva with 4% pooled WMS added was prepared to control for baseline metabolite content of the inoculum. Spectra were analysed by targeted manual quantification of known metabolites. Peak assignments were made using HMDB (hmdb.ca), Chenomx 8.3 (Edmonton, Canada) and literature values. Spectra were integrated into 0.01 ppm buckets from δ 0.7 to 8.5 ppm, excluding δ 4.5 to 5.5 ppm buckets, using MestreC (Santiago de Compostela, Spain), normalised to the standard peak, centred and Pareto scaled and then analysed by principal component analysis and k-means cluster analysis in Knime v.3.4.2 (Konstanz, Germany).

Analysis of bacterial pellets
Microbial pellets were analysed for protein content to ensure protein absent from saliva had not simply been aggregated by bacteria, something that certain oral bacterial species are capable of.
Bacterial pellets were resuspended in 500 µl sterile PBS. A protease inhibitor cocktail (Class I, Sigma) was added as instructed (1% by volume) to prevent any further potential protein degradation during preparation/analysis. Resuspended bacteria were ultrasonically lysed (ten two second pulses, with five seconds in between to prevent heat build-up) and samples of the suspension with and without pellet fragments were prepared for SDS-PAGE as described. An example gel is shown below in Supplementary Figure 2, confirming that there is negligible residual protein in the bacterial pellets and that no protein was lost during pellet preparation. Protein losses from inoculated PS were therefore due to protein catabolism rather than aggregation by bacteria.
Supplementary Figure 2: Coomassie stained polyacrylamide gel showing parotid saliva (lane 1) and the lack of residual proteins within bacterial pellets (lanes 2 -7) or PBS when preparing pellets (lanes 8 -13). Very faint bands can be seen around the amylase bands in the PBS samples, however this does not account for the degree of protein loss observed in incubated inoculated samples. This gel appearance was typical for all samples.

Sensory Scale
An example of the generalised labelled visual analogue scale (glVAS) is shown in Supplementary  Figure 3. Note the sucrose concentration is left blank at the point of rating the scale to reduce rating bias.
Supplementary Figure 3: Example glVAS used for assessing sucrose intensity.

Example 1 H-NMR spectra of PBS and bacteria inoculated parotid saliva
A comparison of the metabolic content of parotid saliva pre-and post-inoculation and incubation with oral bacteria is presented in Supplementary Figure 4. The consumption of host derived urea, citrate and lactate is visible as is the generation of SCFAs, amino acids and phenolic compounds.

Association between salivary protein consumption and metabolite generation by oral bacteria
Supplementary Figure 5: A summary of the significant (p < 0.05) correlations between protein consumption (control lane density minus sample lane density) and change in metabolite concentration. Negative correlation indicates consumption of metabolites whereas positive correlation indicates production of metabolites.
Supplementary Figure 6: Assessment of inter-individual variation of metabolite profiles of inoculated parotid saliva and participants baseline WMS. Following PCA analysis of the samples, variation was assessed by measuring Euclidean distance (weighted for PCA score of the first three axes) between participants for each sample type. A total of fifteen measurements per sample type were made (i.e. all possible pairings from six participants). Mean Euclidean distance was normalised to the anterior tongue samples, which yielded the highest inter-individual variation. ANOVA revealed no significant differences between the inter-individual variation of metabolite profiles for the different sample types.  Table 1: Summary of the concentrations of metabolites consumed and generated following 24 h anaerobic incubation of parotid saliva inoculated with oral bacteria relative to inoculation with sterile PBS. Significant results (p < 0.05) are presented in bold. NA = statistical test could not be conducted due to total consumption of metabolite from all samples yielding a S.D. of zero; n.s. = not significant. Sample means were compared to PBS metabolite concentrations by a one-sample t-test (n = 6).

Metabolite
Post  Table 2: A summary of the salivary concentrations and output changes of salivary metabolites postsucrose exposure, relative to water control. * -glucose is measured as the sum of α-and β-glucose quantified in the sample. Data were analysed by paired t-test (n = 18). Significant p-values (p < 0.05) are included in bold.