'Cyclical Bias' in Microbiome Research Revealed by A Portable Germ-Free Housing System Using Nested Isolation

Abstract

Germ-Free (GF) research has required highly technical pressurized HEPA-ventilation anchored systems for decades. Herein, we validated a GF system that can be easily implemented and portable using Nested Isolation (NesTiso). GF-standards can be achieved housing mice in non-HEPA-static cages, which only need to be nested 'one-cage-inside-another' resembling 'Russian dolls'. After 2 years of monitoring ~100,000 GF-mouse-days, NesTiso showed mice can be maintained GF for life (>1.3 years), with low animal daily-contamination-probability risk (1 every 867 days), allowing the expansion of GF research with unprecedented freedom and mobility. At the cage level, with 23,360 GF cage-days, the probability of having a cage contamination in NesTiso cages opened in biosafety hoods was statistically identical to that of opening cages inside (the 'gold standard') multi-cage pressurized GF isolators. When validating the benefits of using NesTiso in mouse microbiome research, our experiments unexpectedly revealed that the mouse fecal microbiota composition within the 'bedding material' of conventional SPF-cages suffers cyclical selection bias as moist/feces/diet/organic content ('soiledness') increases over time (e.g., favoring microbiome abundances of Bacillales, Burkholderiales, Pseudomonadales; and cultivable Enterococcus faecalis over Lactobacillus murinus and Escherichia coli), which in turn cyclically influences the gut microbiome dynamics of caged mice. Culture 'co-streaking' assays showed that cohoused mice exhibiting different fecal microbiota/hemolytic profiles in clean bedding (high-within-cage individual diversity) 'cyclically and transiently appear identical' (less diverse) as bedding soiledness increases, and recurs. Strategies are proposed to minimize this novel functional form of cyclical bedding-dependent microbiome selection bias.

Figure 1

Nested Isolation System Design. (a) Illustration of ventilation and air filtration in housing systems commercially available for mice, and our Nested isolation system (NesTiso; non-HEPA air filtration occurs in inward/outward directions as air currents move by natural ventilation and external aeration). Mouse photograph, thermography demonstrates mice are a source of heat that instantly affects the temperature of the bedding material and surrounding elements via infrared rays reflectivity. Circles illustrate hottest spot near the eye (35.9 °C), and instant infrared reflection (heat radiation) that warms up surrounding surfaces (e.g., +2.9 °C on bench top; details in Supplementary Figs 1–3). (b) NesTiso setting in ultrabarrier GF room. (c) Germ-free NesTiso cage set in biosafety cabinet housing one 40 week-old GF-mouse during a 7-day DSS experiment (day 72 in NesTiso). Filter lids are sealed to cage bottoms using plastic wrap. Notice the space between the cages to store materials for individualized repeated aseptic handling and weighting of mice (small orange box). (d) Comparison of naturally occurring air humidity inside heavily soiled empty GF mouse cages monitored over time in NesTiso or standard single caging (NesTiso labeled as DC for double caging in illustration, and SC for single caging; 3 cages/group). Notice that NesTiso ventilation dynamics parallel that of SC. Air humidity differences were stable for four days and noticeable immediately after soiled cages were set as NesTiso (y-axis, oval). (e) Effect of external aeration with a household fan on the humidity (wet weight) of experimentally moistened soiled corncob bedding material (replicate sets A and B; without mice). Inset, actual bedding weight in grams (four replicas/cage) over time. Notice markedly improved ventilation and evaporation (bedding desiccation) in both NesTiso and SC. Paired-t test, 4–6-cages/4-replicas/cage.

Figure 2

Nested Isolation has no negative impact on murine phenotypes. (a) The cecum size in NesTiso GF-SAMP mice is significantly larger compared to SPF mice as expected, and remains unaffected across ages (curve slope~0.01, P > 0.1; 3 experiments, triangles vs. circles, t-test P < 0.0001, n = 40). (b) Multivariable unsupervised cluster analysis of SAMP cecum and 8 other organs (normalized biomass, % of body weight) shows NesTiso mice (‘2LNesting’) are identical to mice raised in isolators in the same facility (Isolator.C). SPF and GF-isolator-T mice were included as external comparators. Mice from isolator T (Taconic, Inc.) clustered separately due to lower cecum size after transportation. Correlation statistics predict NesTiso and ‘Isolator.C’ cluster (P < 0.001, see univariate/hematocrit data in Supplementary Fig. 5a,b). (c) 3-D-stereomicroscopic profiling of the small intestinal mucosal surface illustrates the presence and progression of typical ileitis with ‘cobblestone’ lesions in NesTiso GF-SAMP. Scale bar, 1 mm. See histological features of cobblestone ileitis in Supplementary Fig. 5c–e). (d) Right-censored survival analysis (outcome variable: time to death) shows there are no differences on mortality incidence comparing GF mice raised in isolators vs. NesTiso. 72 and 151 day censored data (n = 67 mice, >40-wks old). (e) NesTiso has no negative effect on body weight as surrogate for animal welfare. Twelve-week-old SAMP mice gained or maintained weight as expected while housed for additional 12 weeks in NesTiso (1–2 mice/cage). (f) Quantitative 16s rRNA PCR enumeration of microbial abundance on the feces of a normal human donor and four transplanted NesTiso GF SAMP mice shows relative stability of bacteria families over time (1 mouse sampled per NesTiso set). Inset panel illustrates high qPCR test reproducibility. Abbreviations: Human fec., human feces; F. transplant, day of fecal transplantation; Bifidobact., bifidobacteria; SFBact-1008, segmented filamentous bacteria with 16S rRNA primer R-1008.

Figure 3

Fecal microbiome signatures of donor, transplanted and other mice support NesTiso microbe containment in BSL-2 facility. (a) Microbiome ‘phylum signatures’ profile for replicated samples from the feces of a normal human donor, and below, the corresponding fecal profiles of GF-SAMP mice that received the donor sample as transplant (FMT) on day 1 (10 mice transplanted; 4 NesTiso sets, to diminish within-mouse repeated measures data dependency 8 collected frozen fecal samples were randomly selected for analysis, since individually ventilated cage data has earlier shown appropriate clustering over time). Normalized 16S rRNA gene (microbiome) abundance of fecal bacteria at the phylum level (Y-axis), for all 31 possible phyla in this experiment (X-axis). Notice similar quantitative Firmicutes-rich signatures of donor and FMT mice over time (binary data, 6/6 of 31 possible taxa present), suggesting FMT colonizability and stability in NesTiso. The slight increase in Bacteroidetes on day 21 compared to day 2 and 11 suggests natural enrichment in mice possibly due to different diet and digestive biology compared to humans. (b) Fecal microbiome phylum signatures of SPF B6 mice (used as external comparator) are richer in Bacteroidetes. Binary probability statistics (presence/absence) for all phyla observed in FMT (including low abundant phyla circled in inset plot Fig. 3a on day 21) and SPF mice, indicates that each group had unique signatures and that cage-cage cross-contamination in FMT studies is unlikely using NesTiso (considering the limitations of microbiome data, this data is supported by the 100% prevention of cage-cage dissemination of microbes in NesTiso; see text). (c) Correlation of raw 16S rRNA gene read counts in the feces of the human donor in panel 3a, for two technical aliquots (A and B) to that of the sum of A + B reads (interpreted ‘in series’). Notice that low abundant read data distribution improves the linearity when reads in A and B are added (see diagonal ovals in ‘Order’ panel; Supplementary Fig. 6).

Figure 4

NesTiso does not bias the fecal microbiome, but enrichment of fecal Bacillales and Pseudomonadales in mouse bedding reveals mechanistics of novel form of ‘cyclical microbial bias’ in microbiome research. (a) Split-plot experimental design to assess effect of NesTiso on fecal microbiome. Moist corncob bedding with SPF SAMP-mouse feces was randomly divided into petri dishes, and incubated inside either ‘Single’ or ‘NesTiso’ static cages at 23 °C for 28 days (4 dishes/cage; 4 single vs. 6 NesTiso cages). (b) Box plots with 16S rRNA microbiome read abundance of four bacterial Orders (Clostridiales, Bacteroidales, Actinomycetales, and Lactobacillales) from pooled bedding material shows no difference between Single and NesTiso cages after incubation (n = 4 vs. 6 cages, t-test P > 0.05). Notice consistently reduced data variability (standard deviations) in NesTiso. (c) Multivariate principal component analysis (PCA) of 16S rRNA microbiome bacterial orders illustrate NesTiso does not affect the microbiome of mouse feces in bedding material compared to single caging. (d) Comparative biplot microbiome analysis of mouse bedding and fecal samples illustrates Bacillales and Pseudomonadales [and Burkholderiales in Supplementary Fig. 7] as major orders enriched in bedding material independently of caging type (explaining 58% of data variance, x-axis). Clostridiales and other anaerobes in left side of biplot cannot grow aerobically. (e) Normalized box plot of mouse microbiome data from feces, cecum content, intestinal tissues and bedding samples to contextualize the comparable enrichment of Bacillales in both bedding and intestinal villous samples in mice. (f) Infrared analysis of cages housing conventional SPF-mice illustrates that the temperature in the cage bedding can be as high as 31 °C (mouse max = 37.4; outside cage min = 21.3), which could favor the enrichment of fast growers introducing cyclical selection bias favoring suitable aerobic microbes in mouse microbiome research. (g) Mathematical modeling and simulations over several dilution events using a mechanistic event-customizable model predicts that fecal microbial persistence or extinction in the mouse cage bedding depends on the speed of growth of each fecal microbe in the bedding over cumulative dilutional cage replacement events (set in the model as mouse bedding cage replacements every 10 days; see script and cyclical Periodicity Rules in Supplementary Materials).

Figure 5

Exposure of GF mice to different SPF bedding ‘soiledness’ results in distinct colonization patterns and cyclical bedding-dependent (CyBed) microbiome variability in the mouse gut. (a) Semi-quantitative fecal culture (‘co-streaking’) assay illustrating two distinct fecal culture profiles of five littermate SAMP mice cohoused for 20 weeks. TSA blood agar, aerobic, 37 °C, 5d. See appearance after 36 h of incubation, and follow up gram stain of fecal smear in Supplementary Figs 8 and 9, respectively. Fecal enumeration and single-colony Sanger sequencing indicates abundant cultivable microbes contribute major fractions of bacterial DNA in mouse fecal microbiome. Under the assumption that cultivable and uncultivable microbes interact dynamically, the assay serves to monitor the comparative dynamics of fecal systems. (b) Experimental design to determine the effect of soiledness on colonizability differences in GF-SW mice, and the dynamic effect over three cage replacements. (c) Aerobic incubation of ‘co-streaked’ fecal samples illustrates cultivable microbiota differences. Notice ‘co-streaking’ fecal profiles of 9 SW mice (labels, 1–9): mice look similar on day 1; then appear more distinct with 4 cultivable profiles on day 3; then similar on day 8 (two profiles). Inset line plot, number of ‘co-streaking’ fecal profiles over 33 days (3 bedding cycles). Notice pattern of ‘co-streaking’ fecal profile variability oscillates cyclically over time with every new cage change (more alike when beddings are 10-day-soiled; more distinct when samples are collected three days in clean cages, i.e., 3-day-soiled). (d) Anaerobic hemolytic (virulence) fecal profiles on day 10. Notice that 4 mice exposed to 1-day-soiled bedding have abundant hemolytic anaerobes (absent in 10-day-soiled bedding mice). Exposure to variably soiled bedding affect collective virulence profile of acquired/transmissible microbes from bedding. Because microbiota abundance and virulence variation may influence animal phenotypes, it is necessary to control for CyBeD microbiome variability to improve scientific rigor during experiments, but also during breeding since newborn pups from a single colony may be variably imprinted by the cyclically biased bedding microbiome.

Figure 6

In vitro growth of three abundant fecal aerobic bacteria in bedding shows Enterococcus faecalis resilience to soiledness. (a) Close up photograph of a TSA agar streaked with feces of AKR mouse after 48 h of aerobic incubation 24 h, 37 °C. Notice three abundant distinct colonies (spreading, Escherichia coli; white, Enterococcus faecalis; grey, Lactobacillus murinus) that could be semi-quantitatively ranked and compared. Notice E. faecalis inhibition over L. murinus when in close proximity (dashed circle/inset close-up). Gram-stain morphologies are shown. (b) Schematic of observed interactions among the selected microbes in TSA and other hypothetical uncultivable microbes. (c) Design of in vitro experiment to determine if a mixture of three bacteria could equally grow on bedding at different concentrations of GF-soiled bedding, clean bedding and diet. Inoculated bedding samples were incubated at 23 °C for 9 days. (d) Sanger sequencing chromatograph of bacterial DNA samples from selected and enumerated isolates confirms enumeration data. (e,f) Line plots illustrate that when incubated as a 1:1:1 mixture, E. faecalis is highly resilient to soiledness, and able to readily grow on the GF-grade rodent diet used. Unexpectedly, E. coli was the least adaptable fecal microorganism in the cage environment. Biologically and experimentally relevant, L. murinus, a worldwide aerobic species is best adapted to 5-day-soiled bedding, indicating selection bias favors its abundant growth until it is inhibited by the overgrowth of E. faecalis towards bed-day 10 (Supplementary Fig. 10). These findings derived from SPF-AKR mice confirm the cyclical predictions illustrated in panels 6b-c, which derived from interpretation of the fecal co-streaking profiles of the SPF-SAMP microbiota that was transmisible to GF-SW mice.