A Therapeutic Strategy for nvCJD
(The disease model of Prusiner is assumed below.)CJD has been a untreatable and invariantly fatal disease that can take decades to develop. This slow onset may offer a window of opportunity to develop a therapy to benefit the tens of millions of people now exposed to nvCJD, who in many cases probably already incubating the disease. The BSE epidemic could take 50 years or more to fully play out in the human population; as with prostate cancer, many people may die with it, but not from it.
Clinical symptoms of nvCJD are the end result of a long and complex chain of pathological events, directly and indirectly resulting from exposure to abnormal prion and recruitment by it of normal prion. Intervention might in principle target any stage of the process from exposure to agent, adsorption, migration, conversion of normal prion, cell membrane changes, neuronal cell death, gliosis, or just the symptoms themselves.
While common sense suggests intervening early before massive neuronal cell death occurs, in Alzheimer's (a common CJD misdiagnosis, with similar late onset, amyloidosis, difficulty of timely diagnosis, and impairment of mental processes), the strategy instead has been to treat the symptoms of memory loss and push the problems out past the normal life span of the patient. In large measure, this strategy arises from necessity: there is no way of anticipating who will get AD (besides familial and apo e4 allele predisposition).
The diagnostic situation may improve with nvCJD (though it has not with AD). The 14-3-3 test on spinal fluid could provide confirmation of CJD a few months earlier, but any diagnostic based on neuronal cell death is already late with respect to the pathological process. Tonsil tissue has more potential for earlier identification of individuals in whom the disease process is already underway. It is not ideal for population screening.
Without early diagnosis, we are left with treatment of late symptoms or the dubious proposition of administrating a broad-spectrum neuro-active prophylactic pharmaceutical for decades to a large exposed population, many of whom would never have gone on to acquired nvCJD. Such a drug might emerge from large-scale chemical screens; the situation for CJD is far better than AD for in vitro and animal models of disease.
Some well-known compounds have already shown minor improvements in delaying the onset of scrapie in hamsters. Therapeutic chemicals in this class are not tailored specifically to prion gene or protein; so they could interfere with a broad range of brain processes. Serious but subtle side effects can be anticipated over the long term, and these might be difficult to observe beforehand in animals.
Anti-sense prion gene therapy, while it admirably targets just a single gene, reduces or eliminates production of a gene product. Unfortunately, normal prion function is not known despite some 4,350 studies over 70 years. Despite mild impacts to knock-out mice, remarkable evolutionary conservation of prion sequence implies strong natural selection, and so an important function. It is not settled whether delivery of anti-sense nucleotide inside the brain would be short term, under external regulation, or endogenous and constitutive.
Given this situation, I propose a different therapeutic approach that incorporates at the outset the following features:
For my proposal to work, two assumptions must be valid.
- Prions are specificlly targeted, cross-reaction with other proteins in the brain and side-effects are minimized.
- A single one-shot treatment is needed, not life-long daily medication nor problematic genetic engineering.
- Preventative therapy can be given broadly to exposed populations without diagnosis, as the scale of the nvCJD epidemic emerges.
- Drug development is systematic, nearly certain of success, and applicable to other diseases including AD.
- Normal prion function need not be known.
- Useful information emerges whether or not the strategy is ultimately successful
First, mature prion protein on the outer cell membrane must have a specific binding site for some ligand. Prions need not have enzymatic activity and the ligand need not be a small molecule. (If so, so much the better.) However some ligand must be recognized with tight and specific binding, whether it be in oligomerization, transport, cell-cell recognition, signalling, structure, or ordinary catalysis. It is difficult to think of a protein that does not bind something. Reactive analogues of this ligand are then the desired therapeutic agent.
Second, the current scenario for CJD disease progression must be basically correct: conformational recruitment of normal prion by a rogue prion of abnormal conformation. The key idea is that prion protein has a specific recognition site for itself, i.e., like the vast majority of proteins, it forms an oligomer, for simplicity taken here to be the most common oligomer, the dimer. The energetics of binding are such that the heterodimer (one molecule abnormal, the other normal) converts to homodimer (two molecules abnormal). This explains the oddity of CJD: it doesn't progress in knockout deletions because injected or dietary abnormal conformer has nothing to recruit.
What keeps a dimeric protein from growing? The binding site is occupied or sterically blocked; dimers are self-limiting crystal. From structural studies, we know that dimers are usually head-to-head (rotated180 degrees for reasons given by Monod); they are never head-to-tail (like spoons) because this would allow unlimited growth. Something goes wrong in CJD: the binding site is partially unblocked and growth of oligomer is no longer limited, eventually leading to amyloid deposits. This process is modelled in the figure below.
I propose specifically and covalently blocking these growth points in nascent prion aggregates. There is nothing original here: Antarctic fish can live in salt water whose temperature is below the freezing point of their blood. Potentially lethal ice crystals do form, but the fish have an answer: small peptides that cap off seed crystal growth nucleation points. The trick is to methodically identify and develop a reactive blocking compound in the prion situation.
Note that in familial CJD, there is a steady supply of newly synthesized endogenous abnormal monomer. One-time inactivatation of abnormal conformer might not suffice here. For nvCJD, this is not so: rogue prion was dietary, a process which has largely ceased. If most of the abnormal prion can be inactivated, there would be no new supply, and so little progression of the disease. (However, accumulated amyloid remains.) It would not matter if the therapeutic compund cross-reacted to inactivate normal conformer because it is soon replaced due to the constitutive nature of the gene and rapid turnover of this protein.
So how is this magical ligand to be found? A direct method would be to determine the structure of a prion dimer and key residues comprising its interface. However, lack of solubility in solution of even the normal monomer has been an intractible problem. Only a fragment representing half the normal molecule has been characterized so far: it is missing the GPI anchor, the carbohydrates, the repeat region, and the channel forming toxic core. So directly characterizing an abnormal dimer interface is for the future.
I propose the following systematic method for determing the ligand: trace the gene back to simpler organisms where binding sites and function are easy to determine or likely to be already known, then pushed forward incrementally with reactive and specific ligand analogues.
Human genes did not appear recently, they evolved over time through gene doubling and divergence as a member of a protein superfamily. This is not to say that the biological functions has stayed the same; the method does not require this, only that a binding site survive with approximately the same specificity.
At the present time, the human prion gene appears to be present in a single copy with no close relatives, except for the prion genes in 50 other vertebrates. Note homology searches by many researchers have turned up nothing substantative in yeast, fruit fly, or nematode. However, the problem is trying to take too big a bite -- the chicken and marsupial sequences already show signficant changes in some regions, so the current probe can't identify earlier divergences. In other words, the wrong species were sequenced.
The solution is a graduated incremental approach down the evolutionary ladder, whereby earlier divergent sequences from turtle, frog, fish, shark, amphioxus, etc. are determined until such time as the dynamically revised homology probe can recover a member of the prion gene superfamily in a simple and well-studied experimental organism.
For example, the yeast genome is entirely known; if the function and binding sites aren't already known, they can be rapidly determined. Different domains of prion protein may have different origins. Over 1400 superfamilies are represented in yeast -- probably enough for essentially all mammalian genes.
It doesn't matter if the yeast gene has a different function [ie, is paralogous] as long as binding site specificity is somewhat retained. A reactive analogue is developed and pushed forward up the evolutionary ladder, being modified as needed, until a class of compounds is found that can covalently and specifically block a binding site in human prion protein. It is this class of compound that is varied until all the criteria are met for safe catalytic spoilage of conformational recruitment.
Note that this paradigm is on solid ground if prion protein evolved from an existing ancient gene and has one or more specific ligands. But these are properties of every protein. The procedure does not necessarily yield normal prion function per se, but inevitably there will be clues from the ligand as well as from correlation of prion appearance and change as coordinated with contemporaneous advent of physiological evolved structures.
Finally, the procedure is a general one. Sequence information is easy to come by, but assignment of function is hard. There are many human other genes and diseases that could benefit from this algorithm (e.g., presenilin and Alzheimer's). More systematically, strategic genome libraries are constructed and species are sequenced backwards until such time that an interactive homology tool can identify the superfamily representative in a suitable organism for determining function and binding.
Reader comment is welcome.
[© Copyright 2-11-97 THP @ SBMF; FreshStarttmPrifixtmBobanetm® SBMF; patent rights: D. P. Heiken, Esq.; veterinary apps.to Pfizer Corp.]