I’m interested in coming up with a model that explains how prions replicate and how they cause neurotoxicity, in vivo.  Before diving into any modeling, I want to get some key observed facts on the table.  A model needn’t accord to 100% of observed facts, since not all published observations will necessarily be correct, but an ideal model should be consistent with as many as possible of the observations that have been made.

To try to do as little interpreting as possible, I will refer to things by what was actually measured, regardless of what term the author used. Therefore I will say:

  • PrP-res when proteinase K-resistant PrP was quantified.
  • PrPSc when conformation-dependent immunoassay or CDI [Safar 1998] was used to quantify all misfolded PrP regardless of proteinase K resistance
  • Infectivity when PMCA, cell-based assays or bioassays in animals were used to qunantify infectious titer of prions.

After a scan of the literature, here are some important points.

in vivo findings

1. Prion infectivity plateaus before any neurotoxicity is evident.

This observation is practically as old as the prion hypothesis itself.  Prusiner pointed it out in the seminal paper where he coined the term prion [Prusiner 1982]:

After intracerebral inoculation of hamsters with 107 ID50 (median infectious dose) units, about 102 ID50 units can be recovered in the brain 24 hours later. During the next 50 days the amount of agent in the brain increases to 109 ID50 units. At this time the agent is widely distributed throughout the brain and no regional differences are apparent. The  neuropathology is minimal and the animals exhibit no neurological dysfunction… By 60 to 70 days, vacuolation of neurons and astrogliosis are found throughout the brain, even though the titer of the agent remained constant.

A more recent work [Sandberg 2011] has measured this phenomenon at a fine grain, quantifying infectivity using a cell-based assay [Klohn 2003] at a large number of timepoints in a large number of mice and comparing wild-type mice with those over- and under-expressing PrP.

2. Both total PrP and PrP-res increase throughout the course of prion infection, right up through terminal stages.

Perhaps this has been known for a long time, but it’s hard to Google for the earliest appearance of this fact in the literature.  I noticed it recently in [Moreno 2012 - Fig 2] and it is also clear in [Schulz-Schaeffer 2000 - Fig 3] and [Safar 2005 (ft) - Fig S1].  Safar, who did the most quantitative analysis (using mice expressing ~2x wild-type PrP levels), found that PrP-res accumulated exponentially up to 140 dpi (out of incubation time 149 dpi = timepoint 0.94), the latest timepoint at which he checked.  Curiously, though, PrPSc (as measured by CDI) peaked around 100 dpi and had declined a bit by 140 dpi.

3. Different multimeric sizes of PrP-res exhibit different levels of infectivity.

When large PrP-res fibrils are broken into different mulitmeric species, from monomers to oligomers all the way up to large aggregates, some of these species are more infectious than others [Silveira 2005].  Silveira saw highest infectivity in particles composed of 14-28 PrP molecules, with little infectivitiy in small oligomers (< 5 PrP molecules) and little infectivity in large aggregates.  A caveat about this study is that it started from aggregates and then broke them up using “a variety of detergents and sonication.”  I don’t take it for granted that the results are the same as you’d get if you directly fractioned the PrP-res species that are present in the infected brain.  Still, the point is clear: not all PrP-res (and by implication not all PrPSc) is equally infectious.  It’s a good thing, too, because this offers to resolve the otherwise inherent contradiction between point #1 (infectivity plateaus) and point #2 (PrP-res keeps rising).

4. Brains can degrade large infectious doses in a few days.

This is true even in brains expressing PrP at wild-type levels, as is clear from the [Prusiner 1982] quote above (which was from before knockout or transgenic mice existed):

After intracerebral inoculation of hamsters with 107 ID50 (median infectious dose) units, about 102 ID50 units can be recovered in the brain 24 hours later.

Similar degradation of infectivity was observed in the first PrP knockout mouse [Bueler 1993].  Knockout mice were inoculated with scrapie and then homogenate from those mice was used to inoculate other (PrP-expressing) mice to assay for infectivity:

In the case of Prnp0/0 mice, borderline infectivity was noted at day 4 (presumably due to residual inoculum), and no transmission was noted for 1:10 diluted brain homogenates at 2, 8, 12, and 25 weeks after inoculation.

In other words, knockout mice had degraded almost all of the infectious inoculum after about four days.  (Oddly, Bueler also notes that some infectivity remained in at least one knockout mouse at 20 weeks, but this was only seen when the sample was not heated.  Since the other samples were only being heated to 80°C anyway – not nearly hot enough to inactivate prions – it’s not clear why the heating should matter; later authors have not commented much on this and I don’t know that it’s been replicated).

5. Brains can degrade PrP-res & PrPSc with a half-life of about 1.5 days

This is almost the same point as #3, but what was measured was different.  The studies in #3 measured infectivity using bioassays. By contrast, Safar 2005 (ft) treated Tet-off mice with doxycycline to regulate PrP expression levels, then measured PrP-res and PrPSc using blots and CDI.

When the PrP transgene was turned off at 98 dpi (controls died at 149 dpi, so this is 98/149 = timepoint 0.66) using doxycycline, PrPC expression declined to 40% of its original level within a day, and 5% within a week.  PrP-res likewise declined to about 5% of its pre-doxy levels within about a week [Fig 3].  Let’s plug that number into a half-life exponential decay formula:

N(t) = N0 * ½t/t½ 

10 = 200 * ½7/t½

log2(.05) = -7/t½

t½ = -7/log2(.05) = 1.61 days = 38 hours

Which is slightly inflated because it doesn’t account for the fact that, even with reduced PrPC expression, some PrP-res is still being produced during that week.  Safar comes up with an estimate of 36 hours.  By the way, that’s pretty close to 30h, which is the best estimate of PrP-res half-life from cell culture experiments in ScN2a cells [Peretz 2001 (ft)].

6. PrPC turns over rapidly in the brain, with a half-life under a day.

In the above study, Safar also measured PrPC levels after doxycycline administration using CDI, and concluded that PrPC molecules have a half-life of about 18h, though I couldn’t quite follow the math there.  This is a bit longer than estimates from cell culture with ScN2a cells, which are more like 3-6h [Caughey 1989, Borcheldt 1990 (ft)].

7. PrPC is post-translationally downregulated by ~50% starting about halfway through the incubation period.

A couple of years ago it was discovered that PrP paralog Shadoo is post-translationally downregulated during prion infection [Westaway 2011].  At Prion2013 we learned that this has now been shown for PrPC as well [Mays & Westaway, unpublished].  It is hypothesized to result from a protease response by the cell attempting to degrade accumulated PrPSc.

8. PrPC is upregulated through increased translation rates around the time that neuronal loss begins.

In heterozygous Tg37 mice (~3x wild-type PrP expression levels), eIF2α phosphorylation is reported to occur at 9 wpi out of incubation time 12 wpi = timepoint 0.75, resulting in globally reduced translation rates of most proteins, but paradoxically increased translation rates for mRNAs with multiple 5′UTR open reading frames, including PrP [Moreno 2012].  In homozygous Tg37 mice (~6x wild-type levels) this happened at 7 wpi out of 9 wpi = timepoint 0.77 and in wild-type mice this happened at 16 wpi out of 22 wpi = timepoint 0.73.  In other words, it was pretty much the same relative timepoint regardless of PrP expression level, well within the unit of measurement (weeks post-infection).

This story about PrPC upregulation is slightly in conflict with point #7 above; Charlie Mays told me he did see eIF2α phosphorylation in his mice, but only at terminal stages, and he didn’t see an increase in PrP protein expression.

9. Incubation time is inversely correlated with PrP expression level.

This has been known since the first heterozygous PrP knockout mice [Bueler 1993] and the creation of Tga20 mice overexpressing PrP [Fischer 1996].  The dramatic difference in incubation times according to PrP expression level demands that we have some way of normalizing findings across different mouse models.  That’s why I’ve started referring to “timepoints”, where for instance 60 dpi in a model where controls die at 120 dpi is 60/120 = time 0.50.

But this simple act of division also fails to capture the non-linear nature of how various events in the disease course shift when PrP is overexpressed.  For instance: take bioluminescence from Tg(GFAP-luc) transgenes [Tamguney 2009], a marker of reactive gliosis, the brain’s inflammatory reaction to neuronal distress in prion disease.  With an identical infectious dose of prions, wild-type mice will show a spike in bioluminescence at time .50, while transgenic overexpressers will show the spike at time ~.70 [Kurt Giles, personal communication].  In other words, the symptomatic phase of prion disease appears to be proportionally shorter in mice overexpressing PrP.

Here’s another related point that Armin Giese told me, for which I have yet to dig up the appropriate citation.  Apparently compared to wild-type mice, heterozygous knockout mice can accumulate more PrP-res in their brains before succumbing to illness, while transgenic overexpressers of PrP have accumulated less by the time they succumb.

update 2013-06-12: Armin sent me some citations on this subject. This situation is a bit more complicated than I had thought.  Bueler 1994 (ft) examines PrP-res and infectivity levels in wild-type (Prnp+/+) versus hemizygous knockout (Prnp+/0) mice.  In support of point #2 above, both groups had increases in total PrP over the course of infection.  In the WT mice total PrP increased 3-4 fold, and in the hemizygous mice it increased 4-5 fold, but from a lower baseline.  Putting these together suggests that the WT mice accumulated up to 4x WT PrP levels while the hemizygous mice accumulated up to 0.5*5 = 2.5x WT PrP levels.  PrP-res comprised a maximum of 37% of total PrP in the WT mouse brain and 52%$ in the hemizygote brain.  This suggests that total PrP-res was 4x*.37 = 1.48x in the WT mice and 2.5x*.52 = 1.30x in the hemizygous mice.  So total PrP-res accumulation was still slightly lower in the hemizygotes.  As for infectivity, the WT mice plateaued at 108.6 LD50 units, while the hemizygotes plateaued at about half that much, 108.3 LD50s, though the difference between the two was not statistically significant.  The hemizygotes survived far longer than the WT mice (about 24 weeks vs. 2 to 4 weeks) after reaching these (very similar) levels of PrP-res and infectivity.  Bueler argues this may mean that there is a toxic threshold of PrPSc accumulation which is only barely exceeded in the hemizygotes.  That interpretation may be hard to square with Safar 2005 (ft), since Safar’s mice expressed just 0.2x wild-type levels of PrPC and yet also eventually succumbed to disease.

Fischer 1996 (ft) also examines this issue in WT mice vs. Tg overexpressers.  At the time of terminal illness, PrP-res was 3-10% of total PrP in Tg brains compared to 50% in WT brains.  But the Tg mice (Fischer considered a number of different varieties) expressed 5x – 10x wild-type levels of PrPC; Fischer states that the total PrP-res content of Tg mouse brains at terminal illness was about half that of the WT mice.  (Not all of the raw data are there in the paper for me to re-trace that arithmetic).

Of interest, Fischer also notes that WT mice accumulated up to 5x WT total PrP levels in their brains by the time of terminal illness.  50%*5x = 2.5x of that was PrP-res, and presumably 1x was just normal PrPC expression.  The remaining 1.5x must be accounted for by either (1) protease-sensitive PrPSc accumulation or perhaps (2) increased PrPC translation at terminal stages per Moreno 2012.

10. Prion toxicity is cell-autonomous.

Prion infectivity is obviously (indeed, pretty much tautologically) non-cell-autonomous.  But infected, PrP-expressing grafts within a PrP knockout brain degenerate and spew PrP-res all over the brain without causing any neuronal loss in the non-PrP-expressing tissues [Brandner 1996].

11. Not all PrPSc is PK-resistant.

This should be obvious from the existence of prion diseases that produce little or no PrP-res, namely VPSPr [Zou 2010] as well as fatal familial insomnia, both in humans [Medori 1992 (ft) esp. Fig 2, Brown 1995, Telling 1996] and in mice [Jackson 2009 esp. Fig 2].  But even in prion strains that do produce plenty of PrP-res, a significant fraction of PrPSc is protease-sensitive [Safar 1998].  The only cold hard number I could find on this issue was an estimate of 20-40% of PrPSc in the 263K strain being PK-sensitive [Sanjini 2012].

Of note, Safar 1998 found that the relative PK-resistance of different strains, as measured by GdnHCl ratio, was correlated with incubation time [see esp. Fig 3] – thus more PrP-res means faster disease, perhaps because PrP-res is harder to degrade than PK-sensitive PrPSc.  Indeed, in his later work, Safar 2005 (ft) found that about 15% of RML PrPSc was PK-resistant before the PrP transgene was turned off with doxycycline, and that fraction rose to 30% after 56 more days, while total PrPSc was declining.  That implies that, as you’d expect, PrP-res was degraded more slowly than PK-sensitive PrPSc.

However, the correlation between PK resistance and incubation time is still controversial.  At Prion2013, Joel Watts, speaking about his work with mice expressing bank vole PrP, declared that he sees no correlation at all between GdnHCl ratio and incubation time.

12. Late in disease course, a significant fraction of all PrP in the brain is PrPSc or PrP-res.

Safar 2005 (ft) observed that up to 50% of the PrP in the infected mouse brain was PrPSc, as determined by CDI.  This seems to accord with Moreno 2012, Fig 2a as well – if you eyeball the Western blot in Fig 2a you can estimate that by 10 wpi (timepoint 0.83), about half the total PrP is PK-resistant.

13. PrP mRNA levels remain unchanged throughout prion infection

I can’t find the original citation for this but everyone seems to agree, e.g. [Prusiner 1991, Moreno 2012].

additional cell culture findings

It may also useful to bring in some observations from cell culture for which no equivalent in vivo experiment has been done. These may not perfectly reflect in vivo conditions, but it’s good to at least have some information on these issues.  I’ll number them with “C” in front to indicate “cell culture”.  Right now there’s only one I see as critical.

C1. Only < 3% of PrPC molecules that are synthesized get converted to PrP-res, in ScN2a cells.

Pulse-chase experiments integrating radiolabeled 35S methionine into PrP showed that no more than 3% PrPC molecules synthesized ultimately become PrP-res, in ScN2a cells [Caughey & Raymond 1991 (ft)].  I am not aware of a more recent study that has confirmed this.  It’s hard to know how this translates in vivo.  PrP-res in ScN2a cells is continually diluted by cell division [Ghaemmaghami 2007], so the amount of infectious material available to convert nascent PrPC may not have been as high in Caughey’s cell culture as it would be in the brain in terminal stages.  Still, it seems likely that a large majority of PrPC molecules in the infected brain probably do get synthesized and subsequently degraded without ever converting to PrPSc or PrP-res.

PrP molecules per neuron?

Finally, I think it would be nice to have an order of magnitude estimate of the number of PrP molecules per neuron.  I Googled and couldn’t find anything.  Here’s my best attempt at a back-of-the-envelope.

I’ve seen estimates of PrPC expression being < 1 ug/g of brain tissue [Prusiner 1987].  Full-length PrPC has a molecular weight of 33-35 kDa or kg/mol, so 1 ug = 10e-6 g / (~34000 g/mol) = ~3e-10 mol.  3e-10 mol * 6.022e23 molecules/mol = ~2e14 molecules per gram of brain.  Taking humans as the organism (though Prusiner’s data are surely from rodents), humans have ~20B neurons and our brains weigh ~1.5 kg, so  (2e14 molecules/g * 1500g) / (2e10 neurons) = 1.5e7 or < 15 million PrP molecules per neuron.

Does this sound at all reasonable?  The number of PrP mRNA molecules per neuron has been estimated at just 50 [Kretzschmar 1986], so a steady state concentration of PrPC molecules with a half-life of 18h, you’d need to be translating at a rate of 1.5e7 7.5e6 / (18*3600*50) = ~2.5 protein molecules translated per mRNA per second.  This seems high, considering that pre-modification PrP is 253 amino acids, and translation speeds even in bacteria are only ~40 amino acids / second [Berg 2002].  Also I found one estimate of around 11,000 PrPC molecules per white blood cell in humans [Holada & Vostal 2000 (ft)] and though PrP expression is certainly higher in the brain than in peripheral tissues I’d be surprised if it was 1000x higher than white blood cells.

If anyone knows a citation or has a better back-of-the-envelope calculation for PrPC molecules per neuron, let me know.

I plan to expand and update these lists as I find more things I forgot.  In the meantime this will serve as a starting point for exploring ways to model prion replication and neurotoxicity.