Optimization of conditions for in vitro modelling of subgingival normobiosis and dysbiosis

Modeling subgingival microbiome in health and disease is key to identifying the drivers of dysbiosis and to studying microbiome modulation. Here, we optimize growth conditions of our previously described in vitro subgingival microbiome model. Subgingival plaque samples from healthy and periodontitis subjects were used as inocula to grow normobiotic and dysbiotic microbiomes in MBEC assay plates. Saliva supplemented with 1%, 2%, 3.5%, or 5% (v/v) heat-inactivated human serum was used as a growth medium under shaking or non-shaking conditions. The microbiomes were harvested at 4, 7, 10 or 13 days of growth (384 microbiomes in total) and analyzed by 16S rRNA gene sequencing. Biomass significantly increased as a function of serum concentration and incubation period. Independent of growth conditions, the health- and periodontitis-derived microbiomes clustered separately with their respective inocula. Species richness/diversity slightly increased with time but was adversely affected by higher serum concentrations especially in the periodontitis-derived microbiomes. Microbial dysbiosis increased with time and serum concentration. Porphyromonas and Alloprevotella were substantially enriched in higher serum concentrations at the expense of Streptococcus, Fusobacterium and Prevotella. An increase in Porphyromonas, Bacteroides and Mogibacterium accompanied by a decrease in Prevotella, Catonella, and Gemella were the most prominent changes over time. Shaking had only minor effects. Overall, the health-derived microbiomes grown for 4 days in 1% serum, and periodontitis-derived microbiomes grown for 7 days in 3.5%–5% serum were the most similar to the respective inocula. In conclusion, normobiotic and dysbiostic subgingival microbiomes can be grown reproducibly in saliva supplemented with serum, but time and serum concentration need to be adjusted differently for the health and periodontitis-derived microbiomes to maximize similarity to in vivo inocula. The optimized model could be used to identify drivers of dysbiosis, and to evaluate interventions such as microbiome modulators.