Standardizing Mouse Models to Account for Host–Microbial Interactions

Abstract

In a recent perspective piece, Stappenbeck and Virgin suggest new experimental and reporting guidelines to account for phenotypic impact of host-microbiome interactions and reduce irreproducibility between laboratories. In a recent perspective piece, Stappenbeck and Virgin suggest new experimental and reporting guidelines to account for phenotypic impact of host-microbiome interactions and reduce irreproducibility between laboratories. Citation: Stappenbeck TS, Virgin HW. Accounting for reciprocal host-microbiome interactions in experimental science. Nature 2016; 534: 191–199. The National Institutes of Health has launched an initiative to enhance research rigor and reproducibility. Recent appreciation of the variability of the microbiota between animal facilities and the impact of different microbial communities on the host support the contention that phenotypic differences between laboratories may be due in part to different microbiome/host interactions. In the June 2016 issue of Nature, Stappenbeck and Virgin confront host–microbiome interactions in experimental science and propose changes in current practices that, if implemented, will impact all researchers. Though these changes are needed, they come at increased financial, time and administrative costs. The authors urge investigators to consider experimental subjects as a global metagenome, the sum of the genes from the host and its associated microbiota (bacteria, viruses, fungi, archae, and meiofauna). In such a complex multiorganism, and given bidirectional crosstalk between the host and the microbiota, different phenotypes may sometimes be dominantly dependent on host genetics, the microbiome or the combination of both (see Figure 1). This is true not just for organs that are directly colonized by the microbiota, but also for distant organs such as the heart or brain. Indeed, microbes, their products or the cells they influence may circulate to, and affect, distant sites. This new paradigm challenges interpretations of experiments we thought were simple. For instance, the lack of autoimmune diabetes in MyD88-deficient NOD mice in conventional animal facilities is not due to a dependence on MyD88 for diabetes development. Rather, the lack of MyD88 favors a protective microbiota, as germ-free MyD88-deficient mice develop severe diabetes. The authors identify a number of parameters that are not usually disclosed in manuscripts and that can significantly change the composition of the microbiota, which, in turn, may impact phenotypic traits. These parameters include the genetic background of the mice, ambient temperature (mice experience cold stress at temperatures below 26°C, which are often found in mouse facilities and certainly during transport between facilities and the laboratory), water pH, diet, light–dark cycles, age and therapy. The authors provide direct guidelines to improve result interpretation. The first is the use of littermate controls, which, for deletion of autosomal genes implies breeding heterozygous parents to compare wildtype, heterozygous and knockout progeny. If this is not possible, fecal transfer or co-housing of wildtype and mutant mice can at least partially equilibrate the microbiota between the two strains, but will not reveal cases in which the microbiome dominantly modulates a trait that is also dependent on host genetics. The second is mandated disclosure in scientific reports of the exact settings used for each experiment. This includes diet, water treatment, light cycle, genotype and source of the animals (with genomic microsatellite analysis if the mice are from mixed backgrounds). This, in addition to mandatory uploading of microbiome sequencing data to public databases, would allow for the comparison of phenotypes, genotypes and experimental settings between laboratories. The third guideline is that key experiments be performed in multiple animal facilities and institutions via collaborations, similar to how multicenter clinical trials validate initial pilot findings. The use of isobiotic mice (animals colonized with only a defined set of microbes) has been proposed as a way to eliminate microbiome variability across laboratories, similar to the use of isogenic mice to fix the genetic background. However, although this approach is quite useful for select questions, maintaining such restricted microbial association would be cost-prohibitive for many laboratories and prevent analysis of more complex host–microbial interactions. Thus, the use of control littermates, co-housing mutant and control mice, standardized reporting of all experimental details and mandatory uploading of raw data may be increasingly requested of basic scientists. This could have profound effects on transplant research, and appropriately so, as emerging data suggest that the microbiota modulates alloreactivity. Clinical variability in transplant outcomes may also depend on the metagenome, and clinical studies should bank stool samples in addition to genetic material and immune cells to integrate effects of the metagenome on disease susceptibility, progression and responsiveness to therapy.