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.

People have been studying the relationship between copper and prion protein even since before the prion hypothesis.  Way back in the “scrapie agent” days, it was pointed out that cuprizone, a (toxic) copper chelator, can cause spongiform lesions in the rodent brain that bear some resemblance to those caused by scrapie [Pattison & Jebbett 1971], and that the changes in some metabolic markers are similar as well [Kimberlin 1974].    Of course, phenocopy is no evidence of mechanistic similarity.  But one study found, inconsistently, that in some experiments cuprizone appeared to delay or (with very low inoculation titers) prevent terminal illness in scrapie-infected mice [Kimberlin & Millson 1976].  After that, the issue seems to have largely dropped from view for about two decades.

Copper came back into the spotlight in the prion era when it was found that the octapeptide repeats (PHGGGWGQ) in PrP bind Cu2+ ions, specifically at the histidine (H) residue [Brown 1997].   Since then there have been loads of studies on copper and PrP, more than I’ve managed to read.  PrP binding is highly specific for Cu2+ [Stockel 1998], and this binding appears to involve cooperation between the multiple octapeptide repeats [Viles 1999].  Cu2+ binding stimulates PrP endocytosis [Pauly 1998 (ft)].  It’s been reported that PrPSc does not bind copper [Shaked 2001], and that copper binding can be used as an assay to separate PrP by conformation [Muller 2005].  Copper helps PrPSc ”recover” its infectivity and biochemical properties after reversible denaturation by GdnHCl [McKenzie 1998 (ft)], yet copper inhibits PrPC → PrPSc conversion in PMCA [Orem 2006].  Copper can also induce PrPC to acquire protease resistance and detergent insolubility – two properties normally associated with PrPSc - but in a different conformation than PrPSc [Quaglio 2001].    Copper activates transcription of PrP [Varela-Nallar 2006 (ft), Qin 2009 (ft), Bellingham 2009 (ft)].  Phylogenetics suggests that PrP may have evolved from the ZIP family of metal ion transporters [Schmitt-Ulms 2009].

Mice expressing PrP without the octapeptide repeats have extended incubation times despite high PrP expression levels [Flechsig 2000 (ft)].  Both insertions [Goldfarb 1991 (ft)] and deletions [Beck 2001] of the copper-binding octapeptide repeats cause familial prion disease in humans.  It’s been proposed that the age of onset / repeat length association in the insertional mutations may arise from fundamental differences in copper-binding properties observed in vitro [Stevens 2009].

Based on all the connections between copper and PrP, some authors have been tempted to draw a connection between copper and prion pathology [e.g. Brown 2005].  But this is not an easy case to make.  It’s abundantly clear that prion neurotoxicity arises from a toxic gain of function of PrPSc – the loss of PrPC is perfectly survivable [Bueler 1992, Mallucci 2002], and only PrP-expressing cells experience prion-induced degeneration [Bueler 1993, Brandner 1996].  Moreover, PrPC isn’t even lost in prion disease.  In cell culture, at least, just a small fraction of PrPC converts to PrPSc [Caughey & Raymond 1991 (ft)], and in vivo plenty of PrPC can be found even at the stage of terminal illness [Safar 2005 (ft)].  So any explanations trying to give copper a fundamental role in prion pathology will need to be consistent with gain of function.  I haven’t yet seen a theory that ties this together convincingly.

But given that PrP binds copper, it’s not unreasonable to think that copper levels and copper localization could affect prion conversion and thus the course of prion disease.  And particularly since PrP expression level is so well correlated with incubation time [Bueler 1993, Fischer 1996] and copper affects PrP expression [Varela-Nallar 2006 (ft), Qin 2009 (ft), Bellingham 2009 (ft)], it does seem reasonable to think that copper could have an impact on prion disease incubation time.  This hypothesis has been tested in several in vivo experiments.

In this table I attempt to summarize very succinctly the main findings of these experiments, which I’ll discuss further below.

study treatment expected effect on copper in brain measured effect on copper in brain measured effect on PrP expression in brain effect on survival
Kimberlin & Millson 1976 cuprizone - not measured not measured +5%*
Sigurdsson 2003 (ft) D-penacillamine - -26 to -32% not measured +8%*
Hijazi 2003 CuSO4 + +** +** +14 to 25%
Mitteregger 2009 CuSO4 + not measured not measured +?
Mitteregger 2009 Copper-poor diet - not measured not measured -?
Mitteregger 2009 Copper-rich diet + not measured + ?
Mitteregger 2009 trientine - not measured not measured -?
Canello 2012 CuSO4 + not measured not measured -17%
Siggs 2012 Mutated Atp7a gene ? -60% not measured +12%

* Delay was only observed in some experiments and not others
** Only in cerebellum – no change in copper or PrPC elsewhere in the brain

I have added the data from all of the treatment experiments (i.e. all of these studies except Siggs 2012, which was a genetic manipulation) to the Prion Therapeutic Review.

The fundamental conflict is that both copper supplementation [Hijazi 2003, Mitteregger 2009] and copper chelation [Kimberlin & Millson 1976Sigurdsson 2003 (ft)] are reported to extend survival time in wild-type animals.  It’s conceivable that both could actually be true depending on conditions and doses, etc.  If copper binding to PrPC slows conversion to PrPSc, but more copper also means more PrPC expression, then you’ve got forces pushing in both directions.  However, the non-linearity these forces would need to exhibit in order make all reported results actually be true does not seem parsimonious.  It seems easier to bet that at most one of the two claims – either that copper delays disease or that copper chelation delays disease in vivo – is correct.

If so, which?  The results in four of these studies [Kimberlin & Millson 1976Sigurdsson 2003 (ft), Canello 2012Siggs 2012] are all in general agreement that copper is bad, while the other two [Hijazi 2003Mitteregger 2009] suggest copper is good.

Neither Hijazi nor Mitteregger reports p values for any of their experiments, and indeed, they don’t seem to have done (certainly not reported) any statistical tests at all, even though each uses the word “significant” at some point in their text.  Hijazi’s results (14-25% increase in survival time) seem to be too large in magnitude to possibly fail to be significant, but those experiments used n = 5 hamsters, with a single shared control group for all experiments with three different doses of copper.  In such a design it’s conceivable that one or two control animals happening to die early could have created a false positive for all three treated groups, although the standard deviation on the controls (only 4 days) doesn’t suggest a strong influence of outliers.  Meanwhile, Mitteregger’s conclusion states that “copper delays the onset of prion disease in i.p. infected C57Bl/6 mice significantly, whereas a copper deficiency leads to a reduction of the survival time” but the delay and reduction referred to are only +1% (198.7 vs. 195.8 dpi) and -9% (178.5 vs. 195.8 dpi) respectively.

Canello 2012‘s experiment was in a chimeric mouse/human model of E200K familial Creutzfeldt-Jakob disease [Friedman-Levi 2011], and copper supplementation accelerated disease.  It would be nice to see the effect of copper chelators in this mouse model.  It’s worth noting that Canello finds that mutant E200K PrP binds copper poorly.  This could potentially mean that any beneficial effects of PrP binding to copper are lost in this model, while harmful effects of elevated PrP transcription due to copper are maintained.

Kimberlin & Millson 1976 find an effect in only some experiments, and the nature of the experiment (an attempt to exacerbate scrapie via chelator-induced lesions) makes it hard to interpret any “therapeutic” meaning.

Then there is Siggs 2012, whose results were non-intuitive to me at first.  ATP7A is a copper pump that pumps copper out of the cytosol and into the secretory pathway.  So loss-of-function of ATP7A should cause elevated cytosolic copper.  To my understanding of the nuclear pore complex, ions like Cu2+ can simply diffuse through passively, so elevated cytosolic copper should mean elevated nuclear copper, and therefore higher PrP transcription and shorter incubation times.  Instead Siggs finds extended incubation times.  Siggs doesn’t measure brain PrPC expression per se, but of note, total copper in the brain is reduced 60%. Probably I had been thinking about copper distribution too simplistically at a single-cell level – actually copper comes in through the diet, and because of deficient ATP7A, it gets retained in the cells (and therefore organs) it first encounters such as in the digestive tract, and thus fails to reach the brain in the first place.  That story is more consistent with what Wikipedia has to say about Menkes disease.  So PrPC expression could be reduced in Siggs’ mice not due to subcellular localization of copper in the cytosol, but because overall brain copper is reduced.  But is PrPC expression reduced?  Siggs measured brain PrPC in uninfected mice in a Western blot in Fig 2C and Fig 4B (lane 5 vs. 1) and it looks like if anything it might actually be increased, though no quantification is given.  If PrPC expression wasn’t reduced, what was the mechanism for the 12% increase in survival time?

In sum I find the in vivo survival studies every bit as confusing and contradictory as the in vitro and genetic evidence reviewed earlier in this post.  The only thing we can be fairly sure of is that the effects of copper or chelation thereof, whether therapeutic or detrimental, are not large enough to make this a high priority therapeutic strategy.