introduction

For all that we know about prions, no single clear picture has emerged of how exactly prion infection kills neurons.

There have been some compelling proposals.  Accumulation of misfolded PrP triggers the unfolded protein response, leading to eIF2α phosphorylation and a global inhibition of protein translation in the cell, coupled with increased translation of PrP [Moreno 2012].  Cytosolic PrPSc oligomers block entry of substrates into the 20S proteasome [Kristiansen 2007 (ft), Deriziotis 2011].  The evidence for both of these mechanisms is fairly convincing that they contribute to prion neurotoxicity, though neither seems likely to explain it entirely.  At Prion2013, Corinne Lasmezas also presented evidence that prion infection starves cells of NAD+.  (Lasmezas also has reported evidence for monomeric, alpha-helical PrP as the toxic species [Zhou 2012], but a toxic mechanism has not been published yet).

Then there are a few proposals that are interesting but whose connection to actual prion disease is not yet clear.  After it was found that depletion of mahogunin (MGRN1, an E3 ubiquitin ligase) partially phenocopies prion infection, causing spongiform degeneration [He 2003] it was proposed that cytosolic PrP depletes mahogunin [Chakrabarti & Hedge 2009].  However, neither knocking out nor overexpressing MGRN1 influences the course of prion disease [Silvius 2013, Gunn & Carlson 2013 (ft)].  An interstitial deletion mutant of PrP, ΔCR, consisting of mouse PrP without amino acids 105-125 (equivalent 106-126 in humans), creates nonselective pores in membranes that can kill cells by flooding them with cations [Li 2007, Solomon 2010, Christensen 2010] but we still don’t know if that’s what’s going on in prion disease.

Mechanisms aside, there are still basic questions about what is neurotoxic in prion disease.  The obvious answer is: prions.  Which is to say, misfolded PrP; PrPSc (a term by which I do not imply protease resistance).  However, an alternate proposal is that a special “lethal” form of PrP dubbed PrPL which is either an intermediate or a byproduct in the conversion of PrPC to PrPSc.  This notion was first formulated (though not yet dubbed  PrPL) by [Hill 2000] and is discussed further in [Sandberg 2011].

This post will review the proposed lines of evidence for PrPL.

1. prion infectivity plateaus before neurotoxicity begins

One of the early observations about the kinetics of prion disease was that the titer of prion infectivity in the brain plateaus before any symptoms begin [Prusiner 1982].  This is true even in hemizygous Prnp+/- mice and Tg overexpressers of PrP, all of which, surprisingly, accumulate about the same level of prion infectivity as wild-type mice despite very different levels of PrP expression [Sandberg 2011].  Sandberg calls this “an uncoupling of infectivity and neurotoxicity” and argues that this is evidence that “prions themselves are not neurotoxic”.

Certainly, this plateau phenomenon is evidence that infectivity and neurotoxicity are not identical.  But what Sandberg doesn’t discuss is the ample evidence that levels of PrPSc, rPrPSc and total PrP all continue to rise, right up through terminal illness, even after infectivity has plateaued [Moreno 2012 - Fig 2, Schulz-Schaeffer 2000 - Fig 3, Safar 2005 (ft) - Fig S1, Bueler 1994 (ft) - Fig 3A&B].  This fact is entirely consistent with the simple notion that the neurotoxic species is none other than PrPSc itself.  What the infectivity plateau phenomenon really does is to uncouple PrPSc from infectivity, not to uncouple PrPSc from lethality.  These data would have been better evidence for proposing a special infectious species called PrPInf, rather than a lethal species called PrPL.

The discrepancy between plateauing infectivity and rising PrPSc is easy enough to explain away with some handwaving – perhaps as PrP accumulates, neurons manage to sequester it into larger aggregates that are less infectious (but still at least somewhat toxic).  That would make sense, since smaller oligomers of PrP may be more infectious than larger ones [Silveira 2005] and since aggregates seem to appear later in prion disease [citation needed].  Obviously PrPSc is what’s infectious in prion disease – it may just not all be equally infectious.

2. the neurotoxic threshold of PrPSc accumulation is inversely correlated with PrP expression level

Supposedly, Tg overexpressers of PrP require less PrPSc accumulation to cause clinical disease than wild-type mice, which in turn require less PrPSc than hemizygous mice.

If true, this would mean that PrPC levels are correlated with neurotoxicity, perhaps because the rate of PrPC-PrPSc binding events determines production of PrPL, or alternately just because PrPC acts as a receptor mediating PrPSc toxicity [Aguzzi & Falsig 2012].

However I have so far failed to find strong and consistent evidence in the literature for this claim.  The first detailed study of hemizygous mice [Bueler 1994 (ft)] addressed this issue. Fig 3A&B show total PrP and rPrPSc increase throughout disease in wild-type mice, consistent with what I said under point #1 above.  For hemizygous mice (left panels), it’s less clear – rPrPSc and total PrP both appear to level off between weeks 24 and 36.  According to Bueler’s discussion, they level off fairly early at the same level as that reached in terminal stages by WT mice, and this is taken as evidence that hemizygous mice can tolerate PrPSc for longer.  But if you look directly at Bueler’s Western blots (Fig 3B), it appears that the hemizygous mice have less total PrP and less rPrPSc at terminal illness (36 weeks) than wild-type mice do at terminal illness (20 weeks).

The numbers Bueler gives in the body of the paper also suggest less PrPSc accumulation in hemizygous mice.  I addressed this in my prion kinetics post, so, quoting myself:

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.

All this is fairly consistent with the simple null hypothesis that PrPSc is toxic and that hemizygous mice live longer because they have less of it.

Meanwhile, for Tg overexpressers, we have [Fischer 1996 (ft)].  Fischer states that the total rPrPSc content of Tg mouse brains at terminal illness was about half that of the WT mice, but the raw data aren’t there for us to confirm this.  Thackray 2002 reports that rPrPSc was undetectable at the terminal stage in Tga20 overexpressers inoculated with high doses of prions – but sPrPSc is not quantified.

A more recent source of data on this issue can be had in the comparison between WT, Tg37 het and Tg37 homozygous mice in [Moreno 2012 - Fig 1 & Fig S7].  It’s not clear that the loading amount of protein is identical for all blots; if it is, then the blots certainly don’t support a negative correlation between PrP expression level and rPrPSc at terminal illness – if anything, they would support a positive correlation (the WT quantity of PrP looks less than the Tg37s).  Or, viewed a different way that doesn’t depend on loading quantities, Moreno sees significant synapse loss beginning in WT, Tg37 het and Tg37 hom at 16, 7 and 4 weeks respectively, not far off from the appearance of blot-detectable rPrPSc at 16, 9 and 6 weeks respectively.  All this is consistent with PrPSc being what’s toxic.

3. in certain experimental circumstances, PrPSc can exist in the brain without being neurotoxic

Some authors have rejected the idea that PrPSc itself is simply toxic, on the basis that PrPSc in the absence of GPI-anchored PrPC doesn’t kill neurons.  PrPSc is non-toxic in PrP knockout mice [Bueler 1993].  PrP-expressing grafts in PrP knockout mice produce PrPSc and spew it all over the brain, yet only the neurons expressing PrP degenerate [Brandner 1996].  Knocking out PrP in only neurons using an NFH-Cre system allows glial cells to continue to produce PrPSc in the brain, yet neurodegeneration is reversed [Mallucci 2003].  And it was once reported that anchorless PrP mice produce infectivity without neurodegeneration [Chesebro 2005].  (But it was later found these mice do eventually succumb to disease [Chesebro 2010, Stohr 2011] see shedding post.)

One explanation for these phenomena is that there is a toxic intermediate or byproduct of prion conversion, called PrPL, which kills neurons where prion conversion is occurring.  But there are other explanations too.  Aguzzi, who is not a believer in PrPL, proposes that perhaps PrPC acts as a receptor mediating PrPSc toxicity – much like it does for amyloid beta. This is consistent with the fact that some PrP interstitial deletion mutants are toxic [Solomon 2010].

But unlike amyloid beta oligomers, PrPSc isn’t acutely toxic even in wild-type mice expressing PrPC.  Without exception, prion diseases always take weeks before neurons are impaired.  The simplest explanation for all of these phenomena may be that neurotoxicity requires a cell surface or intracellular accumulation of PrPSc - exposure to extracellular prions is not inherently toxic.  This explains Bueler, Brandner and Mallucci’s results just fine.  It would also match fairly well with many of the toxic mechanism theories (see top of post).

This theory doesn’t explain why anchorless PrP mice or humans with truncating mutations in PrP still get sick – but perhaps that’s as it should be.  These diseases, dubbed fatal transmissible amyloid encephalopathy in mice and cerebral amyloid angiopathy (CAA) in humans, exhibit a completely different pathology than prion diseases, with long incubation times and neuropathology that looks a lot like Alzheimer’s and features no spongiform degeneration.  Given how different the disease is, perhaps the toxic mechanism needn’t be the same as it is in conventional “anchored” prion diseases.

Then there are the observations that first prompted the formulation of the PrPL hypothesis.  Hamster prions don’t usually cause disease in mice, which was originally attributed to a species barrier (or more accurately “transmission barrier”) preventing prion replication.  Then it was discovered that wild-type CD1 mice inoculated with Sc237 hamster prions do produce infectivity (108 LD50/g) but don’t succumb to disease [Hill 2000].  Hill argued that therefore “PrPSc and indeed prions… may not themselves be highly neurotoxic”.  However, this study measured only prion infectivity titers and did not perform any really quantitative measurement of PrPSc or total PrP in the brain.  (rPrPSc is blotted in Figs 1 & 3 but not in any quantitative comparison to mice with terminal disease). Given that infectivity plateaus before neurotoxicity even as PrPSc continues to rise (see point #1 above), it seems parsimonious to suggest that the hamster-mouse transmission barrier slowed down replication in these mice such that the infectivity plateau, but not the neurotoxic threshold of PrPSc accumulation, were reached within their lifetimes.  In fact, as the authors note, the mice did exhibit the beginnings of neurotoxicity, with some amount of spongiform degeneration visible on IHC in Fig 2.

Finally, there is also the instance where mouse passage of a particular human GSS case resulted in a strain of prions that seem to cause amyloid plaque deposition but no neurodegeneration [Piccardo 2007].  PrPSc was not detected in many of these plaques by conventional means including PK digestion, NaPTA precipitation and CDI, and the strain was transmissible only to Tg(P101LL) mice and not wild-type mice.  All of these facts suggest this novel prion stain had a different conformation than other prion strains, one which is perhaps less infectious and less lethal.  This is consistent with not all PrPSc being created equal, but it does not require that there be a separate byproduct or intermediate called PrPL, nor is it explainable by positing PrPC as a receptor for PrPSc toxicity.

4. evidence for PrPSc in subclinical infection being comparable to that in terminal disease

Sandberg 2011 also cites three other studies [Race 2001, Thackray 2002, Thackray 2003] as evidence for PrPL as opposed to the PrPC-as-receptor model favored by Aguzzi.

Race 2001 found that brain homogenate from mice inoculated with hamster prions and exhibiting no detectable rPrPSc could produce rPrPSc and disease in subsequent passages in mice.  This doesn’t really argue for dissociation of PrPSc and neurotoxicity, but rather PrPSc and infectivity.  But even there, the dissociation is probably  just a matter of detection limits.  Bioassays for prion infectivity are very sensitive, and infectivity can be detected at 10-9 dilution, beyond the reach of any Western blot.  (And as in most studies, sPrPSc was not quantified).

Thackray 2002 does find rPrPSc in the brain stem of subclinical animals to be fairly comparable to that terminal animals (Fig 3).  But these mice (both terminal and subclinical) had undetectable levels of rPrPSc in whole brain, while Fig 3 shows only brain stem.  So this may indeed be one bit of evidence against a simple model of PrPSc as being toxic, though to be more convinced I’d need to see a similar pattern reported in whole brain, preferably measured more quantitatively with n > 1 and including sPrPSc.

Thackray 2003 shows similar findings in mice inoculated orally in Fig 1.  Strangely, i.p. inoculated mice also have much more rPrPSc than i.c. inoculated mice, though both are assessed at terminal stages.  One possible explanation is a point George Carlson has made to me, that “terminal stages” are defined by physical symptoms such as ataxia which originate from particular brain regions, so that a prion infection can destroy much of the cortex but not be defined as “terminal” until it affects the brain stem or cerebellum. It is possible, then, that i.p. and oral inoculation result in a different starting point of neuroinvasion compared to i.c. injection, thus allowing more rPrPSc to accumulate in other brain regions before a toxic threshold is reached in the critical region(s) that determine “terminal stages”.

discussion

PrPL is a hypothesized lethal species of PrP created as an intermediate or byproduct of prion conversion.  Here I’ve reviewed four proposed lines of evidence for PrPL as the toxic species in prion disease, and have not found strong evidence or logic to support the PrPL hypothesis.

First, the plateauing of infectivity in prion disease does not dissociate PrPSc from neurotoxicity, but rather, PrPSc from infectivity, as PrPSc continues to rise after infectivity maxes out.  Second, I have not found firm quantitative evidence to support an inverse correlation between PrPC expression level and the amount of PrPSc in the brain at terminal illness.  Third, experimental conditions in which PrPSc is present in the brain without causing neurotoxicity are all consistent with a model in which PrPSc is toxic only when tethered to the plasma membrane.  Fourth, there is not an abundance of evidence to suggest that total PrPSc in subclinical animals equal that in terminal animals, and it is important to account for the fact that terminal disease is largely determined by the brain stem and cerebellum even though PrPSc may also replicate elsewhere.

Instead, I have argued that the available data are largely consistent with a simple model in which only membrane-anchored (and perhaps cytosolic) PrPSc are toxic, while extracellular PrPSc is not toxic – at least, not by the same mechanism, as anchorless PrP causes different, non-spongiform, pathology.  This model is consistent with most reported mechanisms of prion neurotoxicity including translational inhibition [Moreno 2012], proteasome inhibition [Kristiansen 2007 (ft), Deriziotis 2011], mahogunin depletion [Chakrabarti & Hedge 2009] and nonselective pore formation [Solomon 2010].

Indeed, given that PrPSc is produced more rapidly than it can be degraded during prion infection [Safar 2005 (ft)], I question whether we will find a single, monolithic mechanism of neurotoxicity in prion disease.  Perhaps the persistent accumulation of misfolded proteins causes a variety of problems, much as an accumulation of garbage in your house would make it unlivable for a variety of reasons.  It is difficult to observe prion neurotoxicity in cell culture, perhaps in part because cell division dilutes PrPSc, preventing toxic accumulation [Ghaemmaghami 2007].  Hopefully the recent observation of prion toxicity ex vivo in brain slices [Falsig 2012] will make it easier to clarify toxic mechanisms in the future.

Suppose that PrPSc is indeed the toxic species, as I’ve proposed, and that prion diseases are diseases of accumulation.  That doesn’t mean that all PrPSc is created equal.  Though I haven’t found convincing evidence for PrPL as a completely separate intermediate or byproduct of prion conversion, I do see reasons to believe that some PrPSc is more toxic than other PrPSc.

At Prion2013, Joel Watts observed that his 109I BvPrP mice have exceptionally short incubation times – around 35 days for most prion strains on second passage – with almost undetectably little rPrPSc, even in strains that normally produce plenty of rPrPSc.  He speculated as to whether the short incubation time in the absence of much rPrPSc might be because 109I BvPrP selectively produces PrPL.  This is an interesting possibility to be explored.

Even besides Watts’ new work, which isn’t published yet, there is evidence in the literature already to support the idea that some strains of prion are more or less toxic than others.  For instance, the comparison of eight hamster prion strains [Safar 1998, Fig 3b] shows that PrPSc (quantified using CDI and thus including sPrPSc) in the brain at terminal illness can vary by a factor of 4, from about 10 ug/mL in SHa(ME7) and DY, up to 40 ug/mL in ME7-H.  This could be due to strain-specific neurotropism – perhaps ME7-H accumulates first in brain regions that don’t cause ataxia and death – and/or it could be due to DY prions being more toxic than ME7-H prions.

Another example is, in fact, fatal familial insomnia.  It’s well known that FFI produces relatively little rPrPSc in humans [Medori 1992 (ft) esp. Fig 2, Brown 1995] and mice [Jackson 2009 Fig 2] though sometimes a similar amount is found [Telling 1996 Fig 1]. What’s remarkable, though, is that FFI doesn’t seem to produce very large amounts of sPrPSc either.  Sick FFI knock-in mice, even at 2 years of age, actually have less total PrP in their brains than uninfected wild-type controls [Jackson 2009 Fig 2F].  How could there be less?  FFI PrP molecules, particularly unglycosylated ones, are selectively degraded before reaching the cell surface [Petersen 1996 (ft), Ma & Lindquist 2001].  What’s remarkable is that even in advanced disease, when neuropathology and behavioral symptoms are evident, FFI PrP has not accumulated to even 1x wild-type PrP levels in these mice.  This suggests that FFI might be a prion strain that is neurotoxic at particularly low concentrations.  (Note this is just mouse data – as for humans, Medori 1992 (ft)  - Fig 3 shows more total PrP in an FFI patient than in a control, and I haven’t been able to find any other good blots showing FFI versus controls).

In any event, the BvPrP and FFI evidence provide at least some basis for speculating that not all PrPSc is equally lethal, just as not all PrPSc is equally infectious. Whether any therapeutic strategies targeting lethality as opposed to accumulation could be successful, however, remains to be seen.  If PrPSc rises steadily throughout prion disease, as it appears to, then blocking one toxic mechanism may only delay disease until the threshold is reached for another toxic mechanism.