The SALAMANDER project aimed at defining salivary signatures indicative of healthy dietary choices (adherence to a Mediterranean diet: MeDi) with a positive long-term health outcome (protection against type 2 diabetes: T2DM).

In a discovery phase, 214 subjects were selected in two cohorts of elderly people (3-C Bordeaux in France, Seniors-ENRICA in Spain) following a case-control design (cases: T2DM). Saliva samples were obtained and their microbiome, metabolome and proteome compositions were analyzed by ¹H-NMR, 16SrRNA gene analyses and nano-LC MS/MS, respectively. This generated three datasets of 78 metabolites, 450 unique amplicon sequence variants and 246 proteins. Each dataset was first analyzed separately. There were differences in microbial diversity with some food groups defining MeDi. For example, subjects who ate more butter showed increased alpha diversity and differences in the presence of some genera including Prevotella, Porphyrmonas and Fusobacterium. Eating more vegetables resulted in a lower presence of Porphyrmonas, Peptostreptococcaceae and Saccharibacteria. For proteome, three food groups were explained by a model comprising proteins: eating vegetables (cohort + 5 proteins), butter (cohort + 4 proteins) and sweets & pastries (diabetic status, sex + 2 proteins).

The datasets of the discovery phase were also analyzed jointly. The microbiome and metabolome datasets shared much information: 60% of the variability in metabolome was explained by microbiome. In contrast, proteome was largely independent of microbiome (<10% of the variability in proteome explained by microbiome). Biomarkers of T2DM were sought in each cohort separately and in a meta-analysis : 5 proteins, ethanol and Coriobacteriia were significantly different between cases and controls. Interestingly, two of the proteins (Q96DA0, Q96DR5) were also significantly different between high and low consumers of butter. 

As a validation, 121 new Seniors-ENRICA participants were selected. 116 saliva samples were obtained and analyzed with the same methods as in the discovery phase for microbiome and proteome. Ethanol concentration was also measured. Overall, none of the T2DM biomarker candidates identified in the discovery phase (5 proteins, ethanol, Coriobacteriia) were confirmed. Only one protein was near significance both in the meta-analysis of the discovery phase and in the validation phase (gene CTBS).

The validation phase nonetheless provided interesting results. It first confirmed that variability in microbiome did not explain variability in proteome. Second, there were statistically significant differences in both alpha and beta diversities between participants who visit the dentist frequently (at least once a year) or not. This was particulary true for T2DM patients.

Overall, it was not possible to identify a signature of T2DM across both a discovery and a validation phase. However, the results strongly suggest that saliva reflects some aspects of usual diet.

In view of identifying biomarkers of diet and health, we analyzed the microbiome, metabolome and proteome of saliva samples. Two key findings should be pointed out :

- Coordination between the three omics studied was evaluated. We found coordinated profiles of saliva microbiome and metabolome : how individuals are grouped is quite similar considering the profile of either saliva microbiome or metabolome. In contrast, saliva proteome has little information shared with the other two omics. This suggests that identification of biomarkers in saliva may benefit from the combined analysis of proteome on one hand and microbiome or metabolome on the other hand, since these two groups convey complementary information.

- The microbiome diversity and proteome composition were altered by the usual consumption of specific food groups such as vegetables or butter & cream. Although this should be confirmed, this opens the way of utilizing saliva (an easily accessible biofluid) as a source of objective markers of usual diet.