Recently I introduced the glycosylation of prion protein (PrP).  In brief: strain information appears to be encoded exclusively in conformation, with no need for glycosylation  [Bessen 1995Piro 2009], at least in most prion strains.  However, different strains present with different ratios of the three principal glycoforms of PrP – di-, mono- and un-glycosylated, and these ratios are stable over transmission through multiple species of animal [Collinge 1996, Hill 1997] and over propagation in vitro.  A plausible model to explain this phenomenon is the ‘selection model’ [Collinge 1996] namely that different strains – where strain is determined by conformation – have differential abilities to convert PrPC depending on the glycosylation state of PrPC.  Because the glycosylation of PrPC varies by brain region even in uninfected animals [DeArmond 1999], the ‘selection model’ could potentially contribute to explaining the known phenomenon of strain-specific neurotropism, meaning that different strains exhibit their most damaging effects on different brain regions.

It’s not trivial to wrap one’s mind around the ‘selection model’.  The purpose of this post is to present a mathematical model of how it could work.


Cells can produce PrPC in four glycosylation states: diglycosylated, mono181, mono197, and unglycosylated.  Here I’ll present a simple model with nice round large numbers.

Suppose Strain A has a conformation such that it efficiently converts PrPC glycosylated at N181, and inefficiently converts PrPC unglycosylated at this site.  In a unit of time t, it is able to convert 80% of diglycosylated PrPC molecules to which it is exposed, 80% of mono181 molecules, 20% of mono197 and 20% of unglycosylated.

Strain B ‘prefers’ minimal glycosylation.  In a given unit of time it can convert only 20% of diglycosylated molecules, 50% of mono181 and mono197 molecules, and 80% of unglycosylated molecules.

In Brain Region 1, glycoproteins get heavily glycosylated.  Per unit of time, it produces 250 diglycosylated PrPC molecules, and 50 each of mono181, mono197 and unglycosylated.

In Brain Region 2, the PrPC expression level is the same but less glycosylation occurs.  Per unit of time, it produces 50 diglycosylated molecules, 50 each of mono181, mono197 and 250 unglycosylated molecules.

Let’s see how many PrPC molecules will be converted to PrPSc in each brain region in each strain:

prion strain brain region 1 brain region 2
A 80%*250+80%*50+20%*50+20%*50 = 260 PrPSc molecules 80%*50+80%*50+20%*50+20%*250 = 140 PrPSc molecules
B 20%*250+50%*50+50%*50+80%*50 = 140 PrPSc molecules molecules 20%*50+50%*50+50%*50+80%*250 = 260 PrPSc molecules

So if different conformations of PrPSc have different abilities to convert different glycoforms of PrP, and different brain regions produce the four glycoforms in different proportions, then different amounts of PrPSc will be produced in different brain regions by different strains.

Meanwhile, PrP-res accumulation tracks pretty well with the extent of pathology across brain regions – at least this has been reported in CJD [Parchi 1996] and FFI [Cortelli 1997].  So we’d expect that Strain A would devastate Brain Region 1, while leaving Brain Region 2 relatively intact, and the opposite would be true for Strain B.

In other words, glycoform selection provides a potential explanation for strain-specific neurotropism.

does it work with real data?

The numbers used in the toy example above may not be very realistic.  For one, < 3% of PrPC molecules ever get converted to PrPSc, at least in cell culture [Caughey & Raymond 1991 (ft)].  And glycoform ratios in real prion strains are not as dramatically different as the difference between Strain A and Strain B above.  Also, prion strains might “prefer” particular PrPC glycoforms on two different dimensions: which sites are glycosylated (di/mono181/mono197/un) and also what glycan chains are attached.  There is some info on how glycan chains vary by brain region [DeArmond 1999], but I couldn’t find any good data on the ratio of di/mono/un-glycosylated PrPC varies by brain region.

I am thinking of using ImageJ and this tutorial to do densitometric scanning on Western blots of human and rodent PrPSc from different strains [Collinge 1996Somerville 1997] to quantify the glycoform ratio.  If I can also find some data on how the ratio varies for PrPC by brain region, I could actually plug in the numbers and see if they predict the neurotropism observed in these strains.