update 2013-09-22: this post is deprecated.  When I wrote this I was unaware that Jean Manson’s group has already made knock-in mice expressing wild-type human PRNP with both the 129M and 129V alleles [Bishop 2006]. If this can be done, then surely the same could be achieved with a mutant human PRNP allele, though as far as I can tell no one has done it yet.  See this post for a review of humanized mouse models.

Original post below.


After getting my head around the difference between knock-in and transgenic mice, my next question was: would it be possible to introduce a fully human PRNP mutant gene (say, D178N cis 129M = FFI) into mice through knock-in?

I asked this question out of speculation (perhaps unfounded, as we’ll see shortly) that this would be the most faithful possible mouse model. Knock-in models have the disease gene in the right locus and with the right expression pattern, and so can be more faithful than transgenic models.  Meanwhile, fully human PrP might conceivably have a drug response different from mouse PrP, even with the equivalent mutation (say, D178N in humans / D177N in mouse). Some candidate therapeutics have shown different effects in different prion strains (see cpd-B), and if human and mouse PrP are different enough to have a documented species barrier, perhaps they’re different enough to respond differently to small molecule intervention aimed at preventing or delaying prion disease. So if you want to check whether a drug will work in humans, I reasoned, your best bet might be a knock-in of a fully human disease gene into a mouse.

Most genetic mouse models of prion disease use mouse PrP, not human PrP. Jackson 2009‘s FFI knock-in mice use a primarily mouse amino acid sequence with a couple of humanized amino acids.  Most transgenic models use fully murine PrP: for instance the Mastrianni lab’s GSS mice [Yang 2010] or the Ferrari lab’s D178N 129V CJD mice [Dossena 2008].

Some mouse models do use human PrP, but seemingly always as a transgene rather than knock-in. Several investigators have worked with transgenic mice expressing wild-type human PrP (usually 129MM homozygous) and challenged them with intracerebral injections of prions [see for instance Plinston 2011]. Transgenesis has also been used to introduce human PrP with disease-causing mutations into mice: Collinge’s lab at the MRC Prion unit has created mice with human P102L (GSS) or E200K (CJD) transgenes [Asante 2008]. You’d think that these would be pretty excellent mouse models of these human genetic prion diseases, but you’d be wrong: Asante discovered that “no spontaneous disease developed in aged animals”, though in some cases the mice did exhibit heightened susceptibility to challenge with prions from human brains carrying the same mutation. So having the exact human gene is no guarantee the human disease phenotype will translate to a mouse.

So why didn’t he HuPrP E200K mice get spontaneously sick? Was the transgene in the wrong place? Would knock-in have worked? Or is it that human-PRNP-in-mouse not a workable combination? Just last year, the Gabizon lab created a new line of transgenic E199K CJD mice based on “MHu2M” chimeric PrP, meaning part of the amino acid sequence is murine and part is human [Friedman-Levi 2011]. These mice did spontaneously develop prion disease, and quickly too: at age of 5 to 6 months. So why did this model develop spontaneous disease where Asante’s model did not? Friedman-Levi offers some discussion:

While they may be other explanations for the different results in our case, we assume that the introduction of the E200K mutation into a chimeric mouse human PrP, as opposed to a mouse PrP [21] or a human PrP [20], is of biological importance. Chimeric PrP may constitute the bridge that allows human prion diseases to manifest in mice. Indeed, chimeric human mouse PrP was required to transmit at low incubation times genetic and sporadic human prion disease to mice [15]. Moreover, while Tgs expressing the GSS 102 mutation in human PrP did not present spontaneous disease, the same mutation in chimeric PrP did present neurodegenerative disease [28]. Whether the structure of chimeric PrP is more favorable for disease transmission or otherwise the chimeric form has the ability to bind a mouse component important for transmission of human prion diseases to mouse models remains to be established.

So it is possible that the chimeric nature of the protein is important. Mouse and human PrP (often called MoPrP and HuPrP) differ in 28 amino acids. See if you can find them all in the multiple alignment— relative to the human sequence, mice have a one residue deletion, a two residue insertion, and the rest are substitutions.  One important difference is human codons 109 and 112, both M in humans: this completes the seven amino acid 3F4 epitope (KTNMKHM) in humans, a binding site for the monoclonal antibody 3F4 which is often used in prion research.  Mice don’t have the 3F4 epitiope because codons 109 and 112 are L and V instead of M and M.

MHu2M PrP, first developed by Telling 1994, is human at 9 of the 28 sites and murine at the other 19.  Jackson 2009‘s ki-3F4-FFI mice are human at 2 amino acids (in the 3F4 epitope) and murine at the other 26.  There exists yet another PrP chimera which skews more human: the Kitamoto lab has created non-mutant (i.e. no disease-causing mutation) chimeric PrP which they call ChM, entirely human except for a segment at the C-terminus [Taguchi 2003]. Based on Figure 2 this segment appears to have 6 murine amino acids, so the other 22 sites are all human. Taguchi was able to knock this chimera into mice and show transmissibility of some human prion diseases. In the literature searching I’ve done so far, this is the only chimera integrated into mice via knock-in rather than transgenesis. Is that more murine C-terminal segment important in order to have enough sequenece homology for knock-in, or could you do a 100% HuPrP knock-in? As far as I can tell, no one has done it yet. It would be interesting to know whether a human knock-in gene might be a better model than a human transgene, which didn’t work in the E200K example above.

The Gabizon lab’s creation of the E199K MHu2M chimeric mouse model was actually a big step in prion mouse modeling. E200K is the most common mutation causing a genetic prion disease, so people had already tried hard to model it using E199K MoPrP, but according to a review by Weissman 2003 it didn’t work: the E199K mouse had no phenotype. In fact, in that review, Weissman wrote: “the human familial prion diseases have not been modelled successfully in the mouse.” Some genetic models had some amount of neurodegeneration which Weissman called “proteinopathy” but none had yet achieved spontaneous generation of a transmissible disease [Weissman 2003]. Actually, Hsiao 1994 had achieved transmissibility in P102L model mice (with MoPrP P101L transgenes) but only to other MoPrP P101L mice, not to wild-type mice. Jackson 2009 was the first to acheive spontaneous generation of transmissibility to mice without a disease mutation (though still only to Tga20 mice over-expressing PrP, probably because of the 3F4 species barrier).

So modeling prion diseases in the mouse is not a perfected science.  Some models recapitulate the human disease and some fail to do so, for reasons we don’t yet understand.  There have been yet dozens of other attempts that I ran across while searching for papers about this, and no clear answer seems to have emerged as to what factors (location of gene, expression level, specific humanized amino acids) are important for achieving a working model.

update 2012-11-29: a few new facts I dug up. Transmission of FFI prions from human patients to MHu2M chimeric mice was part of an experiment providing some of the early evidence for the prion hypothesis [Prusiner 1997] Later on, Korth 2002 showed that prions from human sFI and FFI patients could be transmitted to MHu2M mice or HuPrP mice but with longer incubation time in the latter.  The chimeric mice were actually more easily infected with the human prions than the HuPrP mice.