Assessment of ecological fidelity of human microbiome-associated mice in observational studies and an interventional trial

Abstract

Composition and function of the gut microbiome are associated with diverse health conditions and treatment responses. Human microbiota-associated (HMA) mouse models are used to establish causal links for these associations but have important limitations. We assessed the fidelity of HMA mouse models in recapitulating ecological responses to a microbial consortium using stools collected from a human clinical trial. HMA mice were generated using different routes of consortium exposure, and their ecological features were compared to human donors by metagenomic sequencing. HMA mice resembled other mice more than their respective human donors in gut microbial composition and function, with taxa including Akkermansia muciniphila and Bacteroides spp. enriched in mouse recipients. A limited repertoire of microbes was able to engraft into HMA mice regardless of route of consortium exposure. In publicly available HMA mouse data sets from four distinct health conditions, we confirmed our observation that a taxonomically restricted set of microbes reproducibly engrafts in HMA mice and observed that stool microbiome composition of HMA mice was more like other mice than their human donor. Our data suggest that HMA mice are limited models for assessing the ecological impact of microbial consortia, with ecological effects in HMA mice being more strongly associated with host species than donor stool ecology or ecological responses to treatment in humans. Comparisons to published studies suggest this may be due to comparatively large host-species effects that overshadow ecological effects of treatments in humans that HMA models aim to recapitulate.IMPORTANCEHMA mice are models that better represent human gut ecology compared to conventional laboratory mice and are commonly used to test the effects of the gut microbiome on disease or treatment response. We evaluated the fidelity of using HMA mice as avatars of ecological response to a human microbial consortium, Microbial Ecosystem Therapeutic 4. Our results show that HMA mice in our cohort and across other published studies are more similar to each other than the human donors or inoculum they are derived from and harbor a taxonomically restricted gut microbiome. These findings highlight the limitations of HMA mice in evaluating the ecological effects of complex human microbiome-targeting interventions, such as microbial consortia.

Keywords: fecal microbiota transplant; gut microbiome; human microbiota-associated mice; microbial consortia.

Fig 1

Different gut microbiota engraft in mouse recipients compared to human donors. (A) Flowchart depicting generation of HMA mice. (B) Principal coordinate analysis (PCoA) plot for Bray-Curtis dissimilarity between human donor and mouse stool composition. Each dot is an individual human donor sample or an average of mouse recipient samples from the same donor. PERMANOVA compares species composition. (C) Bray-Curtis dissimilarity for all comparisons between donors, mice, and between donor and respective recipients. Lines indicate the mean, and whiskers are the standard deviation. Kruskal-Wallis test followed by Dunn’s multiple comparison test was performed. (D) Association between species and relative abundance (RA) of stool taxa in human donors and recipient mice, including donor ID and treatment timepoint as random effects. Blue shading indicates taxa enriched in mice, while red indicates taxa enriched in humans. Statistics and coefficients determined with MaAsLin2. Heat map displays the log % relative abundance of corresponding taxa in individual human and mouse samples. Zero values are arbitrarily set to −3. (E) Scatter plot of delta % RA for stool taxa from baseline to post-MET4 timepoints in human donors and mouse recipients. Each dot represents an individual taxon, and only taxa overlapping between humans and mice were included. Percentage of overlapping taxa is shown for each quadrant. Pearson’s correlation test was performed. **P < 0.01; ****P < 0.0001.

Fig 2

Analysis of the gut microbiome following different treatment exposure routes in germ-free mice. (A) Diagram describing the two routes of treatment exposure in this study. Stools were collected at the indicated timepoints. (B) Principal coordinate analysis (PCoA) plot of Bray-Curtis dissimilarity for MET4, donor stools, and stool from mouse recipients receiving donor FMT and/or direct MET4 administration. Each dot represents an individual human donor sample, an average of MET4 samples sequenced, or an average of mouse recipient stools from the same donor and treatment route. Diamonds represent T0 timepoints; triangles represent T2; and squares represent MET4. PERMANOVA compares MET4, donor stool, and mouse recipient stool composition. (C and D) Average relative abundance (%) histograms for stools from mice receiving post-MET4 (T2) donor stool from B004 (C) and B005 (D) or MET4 directly. Donor stool composition is included as a comparator. Non-MET4 taxa are in grayscale and unlabeled. “Other” consists of taxa with <5% relative abundance in every sample. (E and F) Scatter plots showing log scale % relative abundance of gut taxa in B004 (E) and B005 (F) mice compared between routes of exposure at T2. Each point represents a different taxon. MET4 taxa are represented by green triangles.

Fig 3

HMA mice engraft specific taxa more consistently. (A) Heat map displaying correlation coefficient for relative abundance (RA) of all taxa between studies and species. (B) Comparison of z-transformed correlation coefficients by analysis of variance followed by Tukey’s multiple comparison test. (C) Principal coordinate analysis plot (PCoA) of Bray-Curtis dissimilarity comparing HMA mice to human donors from different studies. Each dot represents a human donor or an average of mouse recipient stools. Circles represent humans; squares represent mice. PERMANOVA compares species composition. (D) Bray-Curtis dissimilarity for all comparisons between donors, mice, and between donor and respective recipients. Kruskal-Wallis test followed by Dunn’s multiple comparison test was performed. (E) Percent (%) engraftment of taxa from human donors into recipient mice in published data sets. Only taxa engrafting in >4 mice are included. Each dot represents % engraftment of one taxon in individual studies, and lines represent the mean across taxa within the study. Taxa were stratified into high, medium, or low engrafters based on average percent engraftment across studies. (F and G) Scatter plots of percent engraftment against average mouse RA across HMA mouse studies (F) or average mouse RA of MET4 taxa in mice exposed to MET4 (G). Each dot represents an individual taxon. MET4 taxa are labeled and are represented by triangles. Colored taxa in panel F correspond to high, medium, or low engraftment from panel E. Lines represent the mean and whiskers represent the standard deviation in all figures. ASD, autism spectrum disorder; CRC, colorectal cancer; IBD, intestinal bowel disease; PD, Parkinson’s disease. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.