The workhorse of mammalian biomedical research is a mouse strain known as C57BL/6. These mice are isogenic: they are all genetically identical. In addition to that, these mice are also homozygous at all sites of their genomes. This is achieved by crossing the mice with each other, brother-to-sister, a process called inbreeding. In an ideal scenario the mice would effectively be clones but in practice there’s still some uniqueness left in each individual (Chebib et al., 2020; Jackson Laboratory 2013). Besides being identical, keeping these strains in a lab for decades induces a selection effect, and thus commonly used lab mice, according to Miller (2016) are far larger, mature much more quickly, have litters about twice the size as those produced by wild mice in laboratory conditions (Miller et al., 2002), are often deaf within their first year of life (Johnson et al., 2000), can no longer make pineal hormones needed for sensing circadian and seasonal cues (Ebihara et al., 1986), and tend to prefer getting into cages than fleeing from them. This occurs because the mice that get to reproduce in captivity are not chosen at random: if a mouse is easier to handle, more docile, and importantly matures fast and has larger litters then it is more likely that those traits will be increasingly represented in later generations.
The commonly given reason behind the pervasive use of inbred mice is to reduce the variance of experiments; this is part of the standard way to think about experiment design. Imagine an experiment where we have two groups of mice, and we give one of them a drug, and we then examine the effects on lifespan by comparing both groups. There is a putative true effect of the drug but there’s also variation in the results due to other factors: genetics (maybe having a particular allele of a particular gene makes a particular type of mouse more responsive to a particular drug), the environment (different temperature, diet, etc.) or an interaction between both. By using inbred mice, it is argued, the variance due to genetics can be eliminated, and by keeping the mice in the same facility and handled in the same way, the variance due to the environment can be reduced, greatly reducing total phenotypic variance. In theory, we are left with just the effect of the drug. In turn, this would seemingly allow for experiments that make use of smaller sample sizes.
While this seems reasonable, it is in fact not true for at least two reasons:
First, this practice can harm the external validity of the study: because the entire population under study is effectively clones of a single individual we may be deriving results that are only applicable to that particular strain. As science journalist Richard Harris puts it in his book Rigor Mortis:
“Imagine that I was testing a new drug to help control nausea in pregnancy, and I suggested to the [Food and Drug Administration (FDA)] that I tested it purely in thirty-five-year-old white women all in one small town in Wisconsin with identical husbands, identical homes, identical diets which I formulate, identical thermostats that I’ve set, and identical IQs. And incidentally they all have the same grandfather.” That would instantly be recognized as a terrible experiment, “but that’s exactly how we do mouse work. And fundamentally that’s why I think we have this enormous failure rate.”
Second and most importantly: the assumption that reducing genetic variance reduces observed (phenotypic) variance is not correct.
Inbred mice are not less variable
Early on during the design of the project, Rich Miller (who runs one of the ITP sites) pointed us to evidence showing that what seems intuitively correct is not so: Empirically, it is not true that one observes less variability when measuring various endpoints for these inbred mice.
Tuttle et al. (2018) aggregated multiple experiments with inbred and outbred mice, pooling over everything from changes in inflammation to femur length or nesting behavior. The result is that Diversity Outbred or DO mice are not more variable when compared to the eight inbred strains that population is derived from:
(From Tuttle et al. (2018))
And broadly, including other outbred and inbred strains and pooling over multiple measurements, there is no obvious difference in variability:
(From Tuttle et al. (2018))
The authors of SLAM (Study of Longitudinal Aging in Mice, using both HET3 and C57BL/6 mice), in their presentation of methods note that the relative variability of outbreds and inbreds is an unresolved question: “Both an inbred and an outbred strain were employed in order to account for genetic heterogeneity as well as strain-specific pathologies. Inbred mice have long been preferred over outbred animals due to the assumption that they may present less phenotypic variability, although this notion has been debated”. Miller (2016) goes further in saying that C57BL/6 (commonly called B6) prevalence may just be a historical artifact: ”when one’s mentor, and her mentor before her, and her mentor’s mentor, used B6 mice, and the reviewers use B6 mice, and the literature is filled with data on B6 mice, and you’ve always used B6 mice, then use of B6 mice and pretending that they are sort of “normal,” in an ill-defined and un-tested way, begins to seem like the right way to do science, just as each lemming is pretty sure that the lemming right ahead must know how to get to the beach.”
All things considered, we agree with Miller here that the weight of the evidence currently available points to either outbred mice being as much or less variable as inbreds. There can be reasons to choose inbred (or F1 mice, the cross of two inbred strains), like experiments involving organ transplantation or simply connecting new research to the bulk of the existing literature, but reduced variability is not one of them.
Why aren’t outbred mice more variable?
At first it seems intuitively obvious that outbred mice ought to be more diverse in their responses to various treatments and their behavior in functional assays due to their diverse genetics. But this obviates the potential role of interactions between genotype and environment.
Phelan & Austad (1994) surveyed the question of outbred vs. inbred variability across a range of species, from tomatoes to mice or chickens, finding that with only a handful of exceptions, outbred individuals present less variability. While certainly variance in the genotype is reduced in inbreds, this general homozygosity means the organism during and after its development has a more limited “toolbox” (set of alleles) than heterozygous mice to converge to a healthy phenotype that is robust to stressors, a phenomenon known as canalization (Flatt. 2005, Voelkl et al., 2020).
Mouse genotype diversity
When Rejuvenome was originally designed, the leading candidate was the UM-HET3 stock (“stock” is used for mice that are not inbred, rather than “strain”). UM-HET3 are mice that are derived from four different parent inbred strains: BALB, C57BL/6, C3H, and DBA. Each mouse in the resulting population is then genetically unique. These are the same mice used by the Interventions Testing Program, as well as the mice chosen for inclusion in SLAM.
(HET3 mice, from Burke et al., 2012)
After various discussions with researchers in the field, we briefly considered so-called “wild-derived” strains: CAST, PWK, and WSB. Wild-derived strains at first sounded ideal: These strains are derived from mice caught in the wild. The origin of CAST lies in mice caught in a barn somewhere in Thailand. But these mice turn out to be also inbred: The original CAST were caught in 1971 but since then, decades of sibling-to-sibling inbreeding have led to a homogeneous population.
CAST, PWK, and WSB are all part of a larger set of mice designed by the Jackson Laboratory (“JAX”) called the Collaborative Cross, which includes those in addition to C57BL/6, 129S1, NOD, and NZO mice. These are all inbred strains, but one can cross them all together to obtain the most diverse commercially available stock of mice: the Diversity Outbred (“DO”) stock which are intended to be as diverse as the human population.
(From Saul et al., 2019)
But between HET3 and DO, which one to choose? We talked to labs that had used both and gathered papers that used them. Being a new stock, there are fewer papers using DOs than there are for HET3s, and some labs told us they had observed that the DOs were particularly “jumpy” and aggressive (in the case of the male mice). But if we could work around the aggression problem, the DOs are clearly more diverse than the HET3. The HETs are always bred from the same mother (CB6F1; itself a cross between a female BALB and a male C57BL/6) and same father (C3D2F1; itself a cross between female C3H and a male DBA), which means mitochondria, which are inherited from the mother, and the Y chromosome, inherited exclusively from the father, are identical in the HET3s. DOs on the other hand do have diversity in their mitochondrial genotype and Y chromosome.
On the other hand, there are decades of data on the HET3 that can serve as reference, and HET3 mice are more reproducible than DOs. A population of HET3s bred in 2000 will have the same allele frequencies as one bred in 2020, but because of the way DOs are bred, there is the potential for drift in the representation of some alleles in the population due to selection effects. This will make it more difficult to compare results between cohorts decades apart. Initially there was an additional reason for DOs (and against HET3s): when Rejuvenome was originally conceived, HET3s could not be purchased directly, they had to be bred in-house by first purchasing the two parent F1 strains. But recently JAX has started offering HET3s and will also offer pre-aged mice in the same way they currently do for C57BL/6, becoming the second available pre-aged mouse strain, which is a welcome development for the field.
An alternative design: diallel cross
We have so far discussed inbred mice and truly outbred stocks where each mouse is unique. But there is an intermediate: crosses between two inbred strains, or F1 mice. These mice are all identical to each other (isogenic), but they are not homozygous at all sites, given that they inherit one chromosome from each parent, and each parent is from a different strain.
One option we discussed early was to cross the 8 founder strains from the Collaborative Cross with C57BL6/J, resulting in a series of mice that are distinct but in a perfectly reproducible way. Doing this could have made it easier to introduce transgenic models (Such as the JAX 011011 mouse that is common in partial reprogramming studies) because these tend to be in a C57BL/6J background. But crosses have the limitation that unlike for HET3 or DO, which can be purchased directly, they have to be bred in house. They also present slightly greater logistic challenges: Two HET3 or DO mice can be treated as fungible, but not so for the F1 panel. Each study design has to make sure to draw the right amount of mice from each of the F1 crosses. They are also less diverse: each mouse is still identical to its siblings. In the end, we decided against any particular cross, in favor of the most diverse mice currently in use in aging research: HET3 or DO.
The choice of mouse strain depends on the research question. For Rejuvenome we are interested in robust data that is not specific to any particular strain, and hence for us outbred mice are the ideal choice. At the moment the exact stock of mice the project will use is not fully decided, but it will be one of HET3 or DO.
We would like to thank Richard Miller for his helpful comments and suggestions
Header image printed with permission from © 2022 The Jackson Laboratory.