A comparison of statin studies in scrapie-infected mice

Read with caution! This post was written during early stages of trying to understand a complex scientific problem, and we didn't get everything right. The original author no longer endorses the content of this post. It is being left online for historical reasons, but read at your own risk.

I’ve blogged previously about the series of experiments demonstrating a therapeutic effect of statin drugs in prion-infected mice.  That post is still useful as a general introduction, but I realized I needed to take a much deeper look, so this post will explore the data from those four experiments in greater detail [Mok 2006, Kempster 2007, Haviv 2008 (ft), Vetrugno 2009 (ft)].

The motivation for this is that, although experimental new drugs of large effect size (like anle138b) are under development, they’re still years away from being prescribed to patients.  Statins are available today, cheap and generic, with an excellent long-term safety record [Law & Rudnicka 2006].  While their effect size is clearly too small to make much difference at the symptomatic stage of prion disease, I am interested in exploring any evidence to suggest whether they might delay disease onset if administered prophylactically in genetic prion disease carriers.

survival data

This table summarizes the survival experiments performed using statins to date:

*This experiment also monitored time to behavioral changes and detected a 13% delay.
**2 mg/kg/day starting, escalating to 20 mg/kg/day.
***Haviv combined the data for these two timepoints because the data were not significantly different from each other.

In all experiments, mice were wild-type, treated orally, and survival figures above reflect time to terminal stages of illness.

Collectively the experiments differed on several dimensions, using two inbred mouse backgrounds (C57BL/6 a.k.a. Black 6, and FVB/N), three prion strains (RML, ME7 and 139A), several different start times for drug administration, both peripheral and intracerebral infection, two statin drugs (simvastatin and pravastatin) and range of doses varying by two orders of magnitude.  Yet the extension of survival achieved by statin administration falls within a fairly narrow range across all of the different experiments, varying from 5% to 12% – only one experiment (Kempster’s 1 mg/kg/day starting at 123 dpi, the onset of behavioral changes) failed to find a significant change in survival time.

That therapeutic effects were found so consistently in very different experimental conditions offers a high level of confidence that the therapeutic effects are real, and not just an accident or an artifact.  Indeed, this level of robustness across so many different experimental conditions in vivo has not been demonstrated for any other candidate prion disease drug, with the possible exception of amphotericin B.  Most of the potential treatments reviewed on this blog have been tested in only one mouse model.

mechanism of action

Since we can be confident that the therapeutic effect is real, it would be nice to know what the therapeutic mechanism is.  Very broadly, do statins inhibit prion infectivity/replication, or prion neurotoxicity?  A fair body of evidence suggests these are not quite identical phenomena.  Infectious titer rises rapidly very early in prion infection, and plateaus long before any neurotoxicity is evident [Prusiner 1982, Sandberg 2011].  I’ll explore the kinetics of prion disease in more detail in a future post.

If statins interfere with prion replication, then they should be more effective the earlier they are administered.  If they only interfere with neurotoxicity, then they’re somewhat more likely to only matter in the endgame, so that early administration doesn’t do much marginal good above and beyond what late administration would do.  That’s not a foregone conclusion, of course – for instance, cholesterol has been estimated to have a 5-year half-life in the human brain [Bjorkhem 1998 (ft)], so if statins reduced neurotoxicity via their cholesterol-lowering effects, then early administration could still be important.

Still, the question of whether statins affect replication, neurotoxicity, or both, is important when we consider whether the ~10% extension of survival in mouse studies might extend to humans carrying genetic prion diseases.  For this class of diseases that often strike at age 40 – 60 [Brown & Mastrianni 2010, Higuma 2013], even a 10% delay in onset, if it could be achieved, would be quite meaningful.  If statins only affect neurotoxicity, then the ~10% delay in onset or survival in mice might only translate to a few months at best in humans, since the duration of the symptomatic stage of these diseases averages 4 – 39 months, depending on the mutation (highest in GSS) [Pocchiari 2004 (ft), see Tables 4, 5, and 6].  On the other hand, if the drugs slow down prion replication or inhibit the spontaneous formation of prions in the first place, then their effects might be appreciable.

It is tempting to infer that, since similar effect sizes were observed in both late (41 – 100 dpi) and early (0 dpi) administration of statins, these drugs must only be interfering with neurotoxicity.  Yet this seemingly simple conclusion fails to present itself clearly in the data.  First, there’s enough variation in mouse survival within each experiment that it’s hard to detect differences in effect size between them.  Moreover, differences in dosage may mask differences in effect size depending on administration time.  Mok’s experiments showing 10% extension of survival with the drug starting at 100 dpi used a dose one or two orders of magnitude larger than Kempster’s and Haviv’s experiments showing 5-12% extension of survival from 0 dpi (100 mg/kg/day vs. 1-20 mg/kg/day).  Vetrugno’s experiment showing a 10% delay at 0 dpi used an even larger dose (200 mg/kg/day) but of pravastatin, a drug generally held to have far lower blood brain barrier permeability than simvastatin [Shepardson 2011, see Fig 2 and Table 3], though Vetrugno cites data disputing this [Johnson-Anuna 2005].  Indeed, the only study that explicitly compared administration at different timepoints under otherwise-identical conditions was Kempster’s study, which indeed found a 5% delay when the drug was given at 0 dpi and no significant effect when the drug was given at the onset of behavioral symptoms.  It does not seem possible to rule out, based on the survival figures, that earlier administration results in larger effect sizes.

Besides survival times, some of these authors present a variety of other evidence trying (largely in vain) to pin down the statins’ mechanism of action.  I summarize these lines of evidence briefly here:

Perhaps the most crucial, and confusing, subject here is the question of whether statins affect PrP^Sc accumulation in the brain. In some experiments, the treated mice and control mice are each examined after they’re sacrificed, which is ~20 days later for the treated mice.  In this comparison, if statins don’t affect PrP^Sc replication (but rather, say, simply allow neurons to tolerate more PrP^Sc before succumbing to neurotoxicity), then the treated mice would be expected to have more PrP^Sc in their brains than controls at time of terminal illness.  If statins do affect PrP^Sc replication but don’t change the threshold of how much PrP^Sc is required for neurotoxicity, then treated and control mice should have identical levels of PrP^Sc at the time of terminal illness.  If statins affect replication and neurotoxicity, then results will be hard to interpret.

Other experiments examine treated and control mice at identical timepoints regardless of disease stage.  In these experiments, if statins affect PrP^Sc replication, then treated mice should have lower PrP^Sc levels than controls.

Authors’ interpretations of their data don’t always match which experiment they did.  Mok compares postmortem brains from treated and control groups, finds no difference in PrP^Sc levels, and incorrectly concludes that this means statins don’t affect prion replication.  On the contrary, Mok’s result suggests statins do affect prion replication, because the treated animals took 16 days longer than controls to reach the same PrP^Sc level.  The question instead becomes one of statistical power.  You can’t prove the negative; did Mok examine enough mice to have power to detect a 10% increase in PrP^Sc in treated mice relative to controls if such an increase existed? It’s not clear.

Kempster, on the other hand, compared treated and control mice at the same timepoint (123 dpi) and found no difference in PrP^Sc.  That result does indeed suggest that statins don’t affect prion replication, but again, it’s not clear that the experiment was powered to detect a difference should one have existed.  Kempster’s figures were 39.3 ± 6.8 ng/mL PrP^Sc in controls vs. 34.9 ± 6.1 in treated animals (n=12, p = .2).  The variance is pretty large to try to detect a small difference with just 12 mice.

Haviv also compared treated and control mice at the same timepoints (105 dpi and again at 130 dpi) and found higher PrP^Sc in treated mice. The presence of higher PrP^Sc in treated mice at the same timepoint is difficult to interpret – Haviv suggests that statins may have inhibited neurotoxicity, thus allowing infected neurons to survive longer and continue producing PrP^Sc.

Vetrugno appears to have, like Mok, compared postmortem brains (i.e. different timepoints) and found no difference in PrP^Sc (“data not shown”) but the methods are not explained in detail.  If true, this result would, again, suggest that statins do affect prion replication.

Although these authors refer to “PrP^Sc“, what they actually measured were changes in PK-resistant PrP (PrP-res), which is but one fraction of PrP^Sc in the brain as measured by conformation-dependent immunoassay (CDI), albeit an important fraction which correlates with disease incubation time [Safar 1998].  There is also no guarantee that statins’ effects on PK-sensitive PrP^Sc were the same as what was measured here.

Nearly two decades ago, it was discovered that statins can inhibit PrP^Sc formation in cell culture [Taraboulos 1995].  The catch was that the drug – in that study, lovastatin – was used at 0.3 uM, while statin concentrations realistically achievable in the brain are more on the order of 1 ng/mL [Wood 2010] which works out to something like 2.5 nM – so about 100 times lower than the concentration used in cell culture.  A second catch was that lovastatin appeared to actually decrease the degradation rate of PrP^C, which could potentially be a pretty bad thing for prion disease.  High throughput assays have repeatedly identified statins as prion formation inhibitors – lovastatin was effective against RML and 22L prions [Kocisko 2003] and lovastatin and rosuvastatin both turned up in UCSF’s PrP-res inhibitor assay with RML prions [Poncet-Montagne 2011], but all of these results were obtained at a concentration of 1 uM. Testing at concentrations of 10, 100 and 500 nM revealed that lovastatin’s IC₅₀ was around 500 nM [Kocisko 2003].  Therefore it’s not clear that the cell culture studies here shed any light on whether statins might be acting via inhibiting PrP^Sc replication in vivo.

In addition to asking about the disease phase at which statins act (replication or neurotoxicity), we can also ask about the specific molecular mechanisms by which they act.  An obvious candidate would be cholesterol levels.  But the three experiments that examined brain cholesterol levels all found it to be unchanged in the treated mice.  Haviv shows evidence that statins act instead through the L-mevalonate pathway, but the significance of this is not clear.  Mok and Haviv both have evidence to suggest increased glial activity in treated mice.  At the end of the day, it’s still entirely unclear what the mechanism of action is.

It’s also unclear whether drug hydrophobicity and blood-brain barrier penetrance are important.  As mentioned above, pravastatin showed similar effects to simvastatin, but at a different dose.  The question of whether BBB permeability matters is important because (as Kempster and Vetrugno point out) simvastatin has a short (~1.5h) half-life [Hess 2001] and is metabolized differently than pravastatin [Klotz 2003, Neuvonen 2008], with implications for dosing and drug interactions.   Even if brain cholesterol levels are important (which these studies do not suggest), Kempster claims that “increasing evidence suggests that reducing peripheral cholesterol levels can affect cholesterol levels in the brain” [citation needed].

Interestingly, epidemiological studies of humans have shown a reduced risk of Alzheimer’s following long-term statin use, and the effect does not appear to depend on the hydrophobicity of the particular statin drug used [Haag 2009].  But no effect was seen with non-statin cholesterol-lowering drugs, suggesting that like in these scrapie studies, cholesterol is not a part of the therapeutic pathway in Alzheimer’s.

As pointed out above, brain cholesterol may have a long half-life [Bjorkhem 1998 (ft)], and though these mice were not treated long enough for cholesterol levels to be affected, longer-run treatment that does affect cholesterol could have additional effects.  After all, we know that PrP^Sc binds to LDLs in vitro [Safar 2006], though we don’t know whether or how this is important.  Brian Appleby has some suggestive evidence that LDL levels may affect disease duration in sCJD, though at present this is still highly speculative.  Half of the patients in his dataset were on statins, pretty consistent with the national average: ~half of men and ~one third of women over 65 take statins in the U.S.  At least, we can rule out the possibility of statins having any really large preventative effect in humans – if they did, sCJD patients should be depleted for statin users.  (To be thorough: Appleby appears to have only included people who had blood lipids drawn, a criterion which ought to enrich for people with known lipid problems and therefore for people on statins, so his ~50% figure could actually represent depletion relative to baseline, but with n=21 it’s impossible to say).

conclusions

In short, little is clear at this point except that statins definitely have some sort of therapeutic effect in scrapie-infected mice.  The mechanism appears to involve L-mevalonate and does not appear to require changes in cholesterol, but it’s hard to say much more than that.

These drugs have never been tested in mouse models of genetic prion diseases, where (unlike in scrapie-infected mice) the disease starts spontaneously from zero infectious titer.  Optimistically, anti-prion drugs may have larger effects when they have an opportunity to inhibit prion replication before it begins.  We shouldn’t be too optimistic, of course – many things work in mice that don’t work in humans, and many of the doses used in these experiments were far larger than any licensed dose in humans.  Simvastatin doses prescribed to humans range from 5 mg to 40 mg/day, roughly corresponding to 1 mg/kg/day (Kempster’s low dose) in mice up to 8 mg/kg/day, still far lower than Mok’s 100 mg/kg/day.

Nonetheless, the effects of statins in prion disease may merit further study.  Unlike experimental new compounds that look to have larger effects (like anle138b), statins are safe, cheap and available today.  Even a small effect on survival would be worth knowing about.