Protein relative abundance patterns associated with sucrose-induced dysbiosis are conserved across taxonomically diverse oral microcosm biofilm models of dental caries

Conservation of function within and between NS and WS pairs

MEGAN5 performs functional analyses by mapping associated RefSeq IDs in BLAST-P output files to functional roles within the SEED, COGs, or KEGG ontologies. COGs is no longer being maintained by NCBI, and most KEGG pathway maps are based on the human genome. SEED is specific for prokaryotes [40], so it seemed the best choice for functional analysis of each metaproteome.

The SEED ontology organizes proteins into metabolic subsystems, which in turn are grouped into major functional categories (carbohydrate metabolism, amino acid metabolism, etc.). Our original intent was to compare SEED assignments for NS and WS microcosms at the subsystem level. However, many proteins were binned into more than one subsystem, an inevitable consequence of overlap between metabolic pathways. This raised the question of whether differences between NS and WS microcosms at the subsystem level might be attributable to a common set of proteins. We also observed that changes in the relative abundance of component proteins within subsystems sometimes went in opposite directions (some were higher in NS, whereas others were elevated in WS).

We decided to analyze the SEED output at the level of individual proteins, in order to facilitate interpretation of the results. Since most spectra were assigned to higher taxonomic levels by MEGAN5 (see above), the findings reported below should be interpreted as a composite picture of responses by taxonomically diverse microcosms to sucrose pulsing.

Additional file 9 provides the complete list of spectral count assignments to SEED proteins. The SEED analysis was able to assign functional roles to 50 % of 1,126,203 total spectra, representing 1969 distinct proteins across all 12 NS and WS pairs. The SEED PCoA plot (Fig. 4) suggested that the NS and WS microcosms were more tightly clustered on the basis of function than the case with the taxonomy PCoA plot in Fig. 3. Almost complete separation was seen along coordinate 1, with the exceptions of 730 NS and 733 NS. The NS microcosms continued to show dispersion along coordinate 2, whereas most WS microcosms were tightly clustered towards the center of that axis (733 NS was an outlier with respect to coordinate 2).

Principal coordinates plot of results for proteins identified by MEGAN5 SEED analysis. The Bray-Curtis distance metric was used. The point labels include the subject number followed by the type of sample. NS microcosms (NS) are shown in green, and WS microcosms (WS) are shown in red. The green and red ellipses are arbitrary and provided solely to aid visualization of each group

Hierarchical clustering results were fully consistent with the spatial distribution of samples in the PCoA plot. In contrast to the results for the HOMINGS and LCA taxonomy analyses, extensive differences in the protein relative abundance profiles for the two major clusters were apparent in the SEED heat map (Additional file 10).

edgeR analysis indicated that 505 proteins (26 % of the 1969 assigned) differed between the NS and WS microcosms at FDR???5 % (Additional file 11). By that criterion, functional analysis was much more successful than either the HOMINGS or LCA taxonomic analyses at identifying differences between the NS and WS microcosms. In other words, sucrose pulsing induced similar changes in protein relative abundance among microcosms that were taxonomically diverse. The following discussion describes major trends, using broadly distributed proteins (present in at least 12 samples), which differed greatly between NS and WS microcosm pairs (FDR???5 %), as representative examples (Table 4). The MetaCyc Metabolic Pathway Database was the primary reference for protein function in NS microcosms [41], except as noted.

The NS relative abundance pattern was consistent with mucin degradation (ABC transporter components for galactose and dipeptides), with mixed acid fermentation directed towards the generation of formate, acetyl CoA, and acetate by various pathways (pyruvate formate-lyase, acetate kinase, acetaldehyde dehydrogenase). Amino acid degradation leading to the release of ammonia was also a prominent feature (NAD-specific glutamate dehydrogenase, lysine 2,3-aminomutase, tryptophanase, histidine ammonia-lyase). All of those proteins were significantly upregulated in NS microcosms at FDR???5 %. Those mechanisms are likely to have contributed to the maintenance of NS microcosms at a stable pH between 6.0 and 7.0. The arginine deiminase pathway has been suggested to participate in ammonification of oral biofilms [13, 42]. Arginine deiminase and carbamate kinase were elevated in NS at FDR???5 %. Ornithine carbamoyltransferase showed a similar trend (FDR?=?0.19).

Our LCA results indicated that S. mutans was undetectable in 8 WS microcosms and a minor component of the remaining 4 WS microcosms (see above). Nevertheless, most studies of oral microbial responses to sucrose and pH stress have focused on S. mutans, due to its status as a presumptive caries pathogen. Accordingly, we used that literature as a knowledge base for interpretation of the WS results (none of the proteins discussed could be unequivocally assigned to S. mutans at the species level).

Sucrose pulsing induced a very distinct relative abundance pattern in WS microcosms sampled at minimum pH on day 3. Multiple sucrose degradation and glycolysis enzymes were elevated at FDR???5 %, notably sucrose-6-phosphate hydrolase, fructokinase, 6-phosphofructokinase, aldolase, and enolase. Aldolase generates fructose-1,6-diphosphate, which activates l-lactate dehydrogenase [42], and both were strongly upregulated (Table 5). This suggested a shift to lactic acid as a fermentation product, which was consistent with the cyclic drops in pH that accompanied sucrose pulsing. Some proteins involved with acid tolerance were upregulated, notably GroES, DnaJ, and components of the F0F1 ATPase. Interestingly, DnaK and GroEL, which are strongly associated with S. mutans acid tolerance [42, 43], failed to meet the FDR???5 % criterion. This may have reflected the low relative abundance of S. mutans in the WS microcosms. In that context, it is worth noting that pyruvate oxidase was strongly upregulated in every WS microcosm. That enzyme is made by most oral streptococci but is absent from S. mutans [44, 45]. Pyruvate oxidase generates H₂O₂, an inhibitor of S. mutans growth in competition experiments [44, 45]. The iron-sulfur cluster assembly protein SufB, which is associated with protection against oxidative stress [46], likewise was strongly upregulated. This would be consistent with the presence of H₂O₂. All WS microcosms did drop below pH 5 at minimum pH, which suggested that non-mutans streptococci played a major role in lactate production.

edgeR results for proteins discussed in the text with FDR???5 % (*): WS vs. NS microcosms

^aNumber of microcosms in which reads were assigned to that protein

Branched-chain amino acid synthesis has been identified as an acid tolerance mechanism in S. mutans [47, 48]. Components of that system were significantly upregulated in all WS microcosms (acetolactate synthase, ketol-acid reductoisomerase, branched-chain amino acid aminotransferase), which suggests that it operates broadly across non-mutans streptococci. Another S. mutans acid response mechanism involves glutamate synthesis [49, 50]. The NADP-specific glutamate dehydrogenase was very strongly upregulated in all WS microcosms. By contrast, the NAD-specific glutamate dehydrogenase predominated in all NS microcosms. Co-factor specificity affects the direction of the reaction. The NAD-specific form degrades glutamate and generates ammonia, whereas the NADP-specific form uses ammonia to synthesize glutamate [51]. Thus, that form has the combined potential to lower the pH via assimilating ammonia, while simultaneously upregulating other acid tolerance systems. Glutamine synthetase likewise was strongly upregulated in all WS microcosms, which further promotes ammonia assimilation. Both enzymes are part of the GlnR regulon, which appears to play a strong role in S. mutans acid tolerance [50]. The WS microcosm results suggest that this mechanism may also be broadly conserved in other streptococci.