Repeated sampling of seven healthy MZ adult twin pairs over a 4-month period emphasized that intrapersonal variation in bacterial community structure was less than interpersonal variation, with co-twins having significantly more similar phylogenetic and taxonomic structure in their fecal microbiota compared to those from unrelated individuals (9
). The results also showed that (i) consumption of a fermented milk product containing 5 bacterial strains was not associated with a statistically significant change in the proportional representation of resident community members within and between individuals; (ii) the appearance and disappearance of strains comprising the FMP consortium did not exhibit familial patterns in the fecal microbiota; and (iii) B. animalis
CNCM I-2494 was the most prominent assayed member of the consortium represented in the microbiota during the 7 week period of FMP consumption. Analyses of the fecal gene repertoire over the course of the 16 weeks of the experiment indicated that (i) variations in the functional features of the (fecal) microbiome were less than the variations in bacterial species composition; (ii) there was no significant difference in the degree of similarity in representation of KEGG orthology group functions for a given co-twin at each time point compared to the degree of similarity that existed between co-twins, while individual and twin pair microbiomes were significantly more similar to one another than those from unrelated individuals; and (iii) there were no statistically significant changes in the representation of these functions when the FMP strain consortium was being consumed. With these findings in mind, and with each individual as well as each genetically identical co-twin serving as a control, we concluded that at least at the depth and frequency of sampling employed for this small healthy cohort, the bacterial species and gene content
of their fecal microbiota/microbiome was not an informative biomarker for understanding whether or how this commercial fermented milk product impacted microbial community properties.
Gnotobiotic mice harboring a model 15-member gut microbial community that represented the three principal bacterial phyla present in the human gut microbiota, and whose 58,399 known or predicted protein-coding genes encompassed many of the prominent functions present in the normal adult human fecal microbiome, provided a means for characterizing the impact of the 5-member FMP strain consortium on expressed gut microbial community functions, and then applying the results to the human fecal specimens collected for this study. As with the MZ twins, introduction of the 5-member strain consortium did not significantly affect the representation of the 15 species comprising the model human microbiota. As with the MZ twins, B. animalis subsp. lactis exhibited the greatest fitness of the five FMP strains in the gut, as judged by its prominence and persistence. Unlike the human arm of the study, where all subjects consumed the FMP twice daily, the design of the mouse study, with its single versus multiple treatment regimens, allowed us to directly compare the persistence of FMP consortium members. Only B. animalis subsp. lactis and L. lactis subsp. cremoris were able to maintain a foothold in the gut ecosystem at detectable levels for the entire 4 week monitoring period after a single dose. In addition, colonization levels were not affected by the number of times the FMP strains were administered to mice.
An advantage of constructing the model human gut microbiome was that its entire predicted gene repertoire was known. This allowed us to define the impact of introducing the FMP strain consortium on the functions expressed by the overall community as well as by its individual components. A major theme emanating from our analysis was the effect of introducing the FMP consortium on carbohydrate metabolism by the community, as well as the effect of the community on a feature of carbohydrate metabolism by B. animalis subsp. lactis. The model 15-member community responded to the FMP consortium by inducing genes encoding enzymes involved in catalyzing reactions that represent the three entry points into the KEGG ‘starch and sucrose metabolic pathway’, as well as enzymes that catalyze fermentation of carbohydrates to propionate. The mechanism by which the FMP strains elicit this response is unclear at present, but the effect is rapid (occurring within the first 24h after invasion) and was persistent whether the consortium was introduced in a single set of gavages during a 1-day period, or with subsequent repeated gavage over a several week period. The persistence of both the carbohydrate pathway response, and of B. animalis subsp. lactis, suggests but does not prove that the latter may be instrumental in instigating and maintaining the former.
Intriguingly, the carbohydrate response showed features of ‘differentiation’. As noted in Results, the levanase response was driven almost entirely by changes in transcription in just a single species (B. vulgatus), the pectinesterase response by 6 community members (B. caccae, B. ovatus, B. thetaiotaomicron, B. vulgatus, B. WH2, C. aerofaciens) and the cellobiose phosphorylase response by three components of the defined model human gut microbiota (B. uniformis, E. rectale, and R. obeum). Of the 50 genes with predicted xylan-degrading capacity in the model microbiome (i.e. those encoding enzymes in ECs 18.104.22.168 and 22.214.171.124), only BACOVA_04387 and BACOVA_04390 (both from B. ovatus) were significantly upregulated after FMP strain introduction (this is ignoring xylanase genes encoded by FMP strains like B. animalis subsp. lactis). This upregulation in a limited subset of the model community coincides with an increase in urinary xylose.
The ability to attribute EC-level changes to individual genes in specific bacterial species was not possible with our RNA-Seq analysis of the human fecal samples. The differentiation of carbohydrate responses among bacterial species documented in gnotobiotic mice emphasizes a challenge and opportunity that can be addressed in these models: namely, to further delineate the niches, interactions and adaptive resource switching behaviors of community members by intentional addition, removal or substitution of taxa, and/or by their modification through genetic manipulation. Although requiring significantly more animals and loss of the ability to use an animal as its own control, future studies could be expanded to include sampling of community gene expression in different segments of the small intestine.
The increased expression of genes encoding enzymes involved in the interconversion of propionate and succinate is intriguing given the fact that this short chain fatty acid has been linked in some reports to effects on gastrointestinal transit time. However, work in this area has yielded varying results and conclusions, perhaps because of the diversity of models and methodological approaches used (27
). Propionate may also link the gut microbiota and human physiology through its effects on hepatic and adipose tissue metabolism (31
). Notably, another group has reported that in the T-bet−/−Rag2−/−
mouse model of colitis, consumption of a fermented milk product containing a dairy matrix plus the same strains used in this study led to increased cecal propionate levels and a reduction in intestinal inflammation (32
The extent of translatability of data from gnotobiotic mouse models harboring collections of sequenced representatives of the human gut microbiota to humans themselves needs to be tested further, not only at the transcriptional level but also at the level of community-host co-metabolism. Although current models can and should be evolved to embrace more of the diversity present in our gut communities, even with current limitations they can serve as part of a pre-clinical discovery pipeline designed to identify candidate biomarkers and mediators of the effects of existing or new probiotic strains on the properties of microbial communities and their hosts. They also represent an analytical tool for characterizing the effects of specified dietary components on the indigenous gut community and on probiotic species that are deliberately consumed. The results could yield new candidate prebiotics that may impact the representation and metabolic properties of probiotic species or entrenched members of our gut microbiota and provide the proof-of-mechanism and proof-of-principle observations needed to justify, direct and interpret human studies.