As noted in our very first post about clinical trials for Fragile X Syndrome:
“The intention here is to take a slightly different perspective.
First, it is interesting to note that in many articles describing the results of clinical trials, the repeated positive results of drug tests in “Fragile X” (FX) animal models are cited in a perfunctory fashion, generally without much, if any, attempt to make a detailed comparison between what was done with the animal and the human subjects.
Second, there is not much discussion of limitations in what animal models can predict about human testing involving higher order cognition. These are two of the issues we wish to address.”
Also noted in that first post regarding animal models more specifically:
“A separate series of posts [beginning with this post] will present comments on preclinical animal research of note to us. In particular, in order to make comparison to the human FXS clinical trials, information will be presented on FX mouse model dosing, phenotypes, and issues previously uncovered (tolerance, assay specificity, etc).”
We are getting close to releasing a first mouse model that we believe should be allow more accurate preclinical testing of neuropharmaceuticals for Fragile X Syndrome. No spoiler alerts here; rather, we will present some of the history as I have experienced it with Fragile X mouse models over the past nearly twenty years. Along the way, it is hoped that some useful principles about the value and limitations of mouse models will become apparent. We will also be presenting data that should make very clear in quite granular way some of the issues discussed.
A personalized history of Fragile X mouse modeling
I will start with my introduction to Fragile X mouse models when I first got involved with the field almost twenty years ago. It very quickly became clear to me, from the first Fragile X conference that I attended, that the mouse phenotypic testing results varied remarkably among the various laboratories in the field. Even more challenging was that the variation in cognitive phenotypes ranged from “mild” to non-existent. This made it somewhat difficult to consider just what phenotypes on which to focus in trying any sort of corrective complementation experiments.
In my case, the FRAXA Research Foundation asked me to become involved because of my expertise in inserting very large segments of human chromosomes into mice. The point then, of course, was to obtain and then insert a region of the human X chromosome into Fragile X model mice to see if normal functions could be restored. Attempts had been made to do this with cDNAs, but there had apparently been toxic effects based on the high and perhaps improperly regulated dosages of FMRP (Fragile X protein product produced by the Fragile X gene, FMR1) that were produced in cells containing the FMR1 cDNAs.
I then had the good fortune to hear a talk at the International Behavioural and Neural Genetics Society (IBANGS) by scientist Robert Gerlai, who was emphatic in his view that mouse researchers were losing sight of “neighbor” effects, i.e. the effects of alleles of genes neighboring the gene of interest being modified in a mouse that were also being retained in the process of making new mouse models.
As we in the field already knew, making “knockout” mice, for example those which had a Fragile X (Fmr1) gene inactivated by researchers, only worked well in certain inbred strains of mice, i.e. primarily strain 129. However, behavioral studies had been developed in other inbred strains of mice, e.g. C57Bl6. Furthermore, making transgenic mice was also biased as to strains which had fertilized oocytes with pronuclei that were easiest to microinject with DNA and which otherwise gave the best recoveries of desired mice, e.g. FVB/NJ. As someone who was had been microinjecting large DNAs into fertilized mouse eggs, I had been using the FVB/NJ strain, and I very much hoped to continue to do so with the experiments I intended to do in the Fragile X field.
What Gerlai was pointing out was that in making knockouts in strain 129, and then backcrossing those mice into the mouse strain of interest, not only would the “knockout” allele of interest (in this case Fragile X disrupted gene) get transferred to the new strain, but so would those 129 gene alleles that were relatively close on the 129 chromosome to the FX disrupted gene allele.
That could be important since mouse strains, like individual people, varied in the functionality of their particular genes. So if strain 129 had 100 genes around the FX disrupted gene, and 25 of those were themselves inactive, those same 100 genes in the desired mouse background, e.g. FVB/NJ, might have a different 25 genes which were inactive. Thus, it would be as if the FX knockout allele was not the only “knockout” (inactive) allele being introduced into the target strain.
Furthermore, the 129 alleles surrounding the FX knockout allele would tend to hang around, since they were close enough that recombination during backcrossing to the mouse strain of interest (in my case FVB/NJ), would be much less likely to quickly exchange them for the FVB alleles of interest. Of course, additional backcrossing to the FVB/NJ strain would eventually get the recombination events to make a “congenic” mouse, i.e. one that really only differed at Fmr1. (A little genetic typography: mouse alleles are written with small case letters after the first capital; human gene names are distinguished by being all capitals: mouse Fmr1; human FMR1; gene names are italicized, while the names of proteins they produce are not. Mutant alleles are written in all lower case, e.g. mouse fmr1.)
Most researchers I met in the Fragile X field were not trained in mouse genetics, so one of the first things I learned was that not knowing the detailed activity of alleles in a particular strain could have a enormous effect on the correct interpretation of research results. My first collaborator in the field, for instance, was studying the effects of the Fragile X knockout on the visual field in the brains of mice. I mentioned this one evening to a colleague of mine at Columbia University, who himself was working on genes which influenced vision in mice. When I mentioned the strain of interest, FVB/NJ, he pointed out that it had a retinal degeneration allele (homozygous, of course, since research mouse strains tend to be inbred). In other words, the FVB/NJ mice became essentially blind early in life. Such a process could itself introduce variation in the way in which neurons in the brain appeared and behaved. This news came as quite a revelation to my collaborator.
In response, the same collaborator then had his laboratory breed the fmr1 knockout mice to sighted mice. While doing so, they selected for elimination of the retinal degeneration genes. This was seeming all very useful, since at the same time they were presumably breeding out the 129 genes around the Fragile X knockout allele. The mice produced by his laboratory were backcrossed for 11 generations, and distributed to other scientists through the Jackson Laboratories.
As it turned out, however, I would learn years later based on experiments which will be presented in subsequent posts/publications, that my collaborator’s lab had not taken into account another issue during backcrossing which greatly impacted what was thought at the time to be the likely status of the region around the Fragile X knockout allele. We will show the data on what we found in a later post, but for now, suffice it to say, that since these mice were among the most highly used in the Fragile X field, it is possible that some variation and discrepancy in results may have had this issue as it cause.
Nevertheless, in the meantime, I resolved to try another approach to reduce the risk from unknown recessive (often inactive) gene alleles in my inbred strain of interest. Pedigrees had been produced for the relationship among the various inbred mouse lines being used in research at the time. I chose a second inbred line, C57Bl6J, which was also of interest in behavioral studies, to breed to the FVB/NJ transgenic lines I had created, i.e. those mice containing one, two, or three copies of the human Fmr1 genomic locus. I had also bred my FMR1 transgenic lines to a mouse line already containing the Fragile X knockout allele. The idea was that in many cases, C57 and FVB strains might differ in the genes which were completely recessive, so that their offspring, called “F1 hybrids”, would have fewer inactive alleles than pure inbred strains, and therefore might have more consistency in behavioral effects that could be attributed to the Fragile X knockout. Furthermore, the human FMR1 transgene was also intended to allow us to “prove” what behavioral effects really came from the mouse Fmr1 knockout, since the genes around the FMR1 transgene, spanning only 90,000 bases (“90 kb”), were not an issue compared to the megabases of neighboring DNA that would accompany a knockout allele during backcrossing.
In order to conduct enough behavioral tests in order to find those that would be strongest in the F1 hybrids, I enlisted the help of an expert in performing batteries of mouse behavioral tests, Dr. Richard Brown, who at that time was the chairman of the Department of Psychology at Dalhousie University.
With the above introduction in mind, the next post will discuss various findings we made relative to strain variation in Fragile X phenotypes.