Prana Biotechnology Limited (NASDAQ:PRAN) is expected to release the results from its Phase 2b IMAGINE trial of PBT2 for the treatment of Alzheimer's disease within the coming month. Immediately prior to the results announcement from the Reach2HD trial, I discussed the scientific basis for believing that PBT2 addresses the pathophysiology underlying Alzheimer's disease. Although the Reach2HD results were thrilling inasmuch as they offered the possibility that PBT2 might have efficacy, they were distinctly underwhelming compared to the dazzling results seen in mice. The best explanation I have for the discrepancy between the animal and human results in Huntington's disease is that Prana is underdosing patients, possibly by as much as a factor of 10. As a consequence, we are likely to see an ambiguous result from the IMAGINE trial. Existing shareholders will be diluted, as the company needs to raise capital or enter partnerships to fund subsequent trials. My estimate of pre-result risk-adjusted net present value is now reduced from $9.60/share to $4.80-$7.20.
The preclinical evidence supporting PBT2 is extremely strong (reviewed in modest depth in my prior article), so I will discuss only the highlights today. Solution-phase data from the related compound clioquinol indicates a square planar configuration of Cu(II) species complexed with clioquinol, with the quinoline nitrogen and hydroxyl groups serving as electron donors. Furthermore, while clioquinol was soluble in water, the copper-clioquinol complexes were only minimally soluble in aqueous solution but were easily dissolved in dimethylsulfoxide, indicating that the clioquinol-metal complexes have increased hydrophobicity compared to the separated organic compound and metal ion. Due to the structural similarity between PBT2 and clioquinol, PBT2-metal interactions should be near-identical between the compounds, except for minor differences in Lewis basicity, binding constant, and hydrophobicity. Binding of hydroxyquinoline derivatives like clioquinol and PBT2 reduces the energy barrier for the passage of metals across the hydrophobic cell membrane.
The biochemical data underlying the metal dyshomeostasis hypothesis of Alzheimer's disease is extensive and compelling:
- Zinc promotes formation of toxic Abeta oligomers
- Zinc and copper are found in high concentrations in senile plaques
- The distribution of zinc-enriched axon terminals corresponds to the pattern of amyloid deposition in the brain
- Alzheimer's disease is associated with alterations in brain zinc and copper levels
- Gene expression for zinc transporters, especially ZnT3, is increased in the vicinity of senile plaques
- Abeta oligomers catalyze the formation of reactive oxygen species in the presence of copper
- PBT2 can extract zinc from senile plaques and deliver it to the intracellular space
- PBT2 promotes clearance of Abeta by a matrix metalloprotease-dependent mechanism
- PBT2 increases the phosphorylation of glycogen synthase kinase 3, which is associated with decreased hyperphosphorylation of tau
While many Alzheimer's researchers do not agree with the metal dyshomeostasis theory of Alzheimer's disease, I consider it to be the best available explanation of the data. PBT2 is the ultimate test of the metal dyshomeostasis hypothesis: if the hypothesis is wrong, why has PBT2 shown efficacy in animal models of Huntington's disease and Alzheimer's disease? It would be serendipitous, indeed, if PBT2 and clioquinol also coincidentally inhibited beta-secretase, promoted the phosphorylation of glycogen synthase kinase 3, or had some other beneficial effect in mouse models of neurodegeneration. The parsimonious explanation of PBT2's activity in mice is that the hypothesis is correct. The whole question is how to run the clinical trials such that differences between treatment group and controls can be detected. This task is not at all easy, given the many unknowns in exploring a novel hypothesis, the limits of experimental manipulation possible when dealing with real patients, and the wide error bars inherent in cognitive testing.
Dosage matters. It would be incorrect to infer that warfarin is ineffective because the prothrombin time failed to change at 0.5 mg/day, or that ibuprofen fails to relieve pain because symptoms did not improve at 20 mg. I believe Prana is underdosing patients.
Ionophores can be dangerous at high doses. Some concern that the hydroxyquinoline ionophores may potentially have severe toxicities stem from the Japanese clioquinol disaster in the 1960s, when 10,000 cases of subacute myelo-optic neuropathy were associated with clioquinol use as an antibiotic, reviewed here. Subsequently, the relationship between clioquinol use and subacute myelo-optic neuropathy has been questioned and it has been suggested that underlying vitamin deficiencies were responsible, rather than clioquinol. Nevertheless, at high doses, clioquinol is definitely neurotoxic in multiple animal species. Typical toxic doses are greater than 200 mg/kg, depending on the species and dosing schedule. Baboons tolerate 28 weeks of 200 mg/kg, but develop obvious signs of toxicity at doses of 600 mg/kg or higher. Although neurotoxicity has not been observed in PBT2, the question remains as to what comprises the safe upper limit. I do not have access to the toxicology data Prana submitted to the FDA.
In the Phase 2 trials utilizing PBT2, doses have ranged from 50-250 mg/day. Huntington's disease patients tend to lose weight during the course of their disease, so the "average person" weight of around 70 kg is a reasonable estimate, yielding doses of 0.7-3.7 mg/kg. For patients with early Alzheimer's disease, who are older and heavier (at least early in the disease), average patient weights will be in the range of 80-100 kg, for dose estimates of 0.5-3.1 mg/kg. IMAGINE is using 250 mg PO qday dosing, roughly 3 mg/kg, with the lighter patients, such as frail elderly women perhaps approaching 4 mg/kg.
In multiple models of both Huntington's disease and Alzheimer's disease, PBT2 and clioquinol have demonstrated efficacy. The Huntington's disease data is considerably sparser, with the best exposition in Cherny 2012, who examined the effects of PBT2 on a worm model (C. elegans) and on a mouse model. The C. elegans model consisted of worms expressing yellow fluorescent protein fused to a polyglutamine repeat. Normally, these worms develop insoluble aggregates of yellow fluorescent protein and eventually become paralyzed. When cultured on nematode growth medium containing PBT2 at 10 micrograms/ml (36.9 micromolar), the YFP-polyglutamine worms still developed the insoluble aggregates, but became paralyzed at an older age than the control worms. The concentration of PBT2 within the worms was not measured.
In R6/2 mice, which bear the N-terminal portion of human huntingtin with an expanded polyglutamine repeat (~120 CAG repeats) and which emulate the human phenotype, treatment with PBT2 prolonged life, reduced weight loss, improved motor function, and decreased brain atrophy. Wild-type mice treated were unaffected by treatment with PBT2. Dosing in the R6/2 mouse model of Huntington's disease was 30 mg/kg.
While in vitro incubation of homogenized human brain from Alzheimer's patients with 0.4 micromolar clioquinol solubilized Abeta 1-40 and Abeta 1-42 quite well, peak activity was seen in the range of 4-28 micromolar (Cherny 2001). In Tg2576 mice, which overexpress a mutant amyloid precursor protein found in familial Alzheimer's disease, a clear dose dependence of clioquinol was identified. Cherny found a nonsignificant decrease in Abeta in Tg2576 mice treated with 2 mg/kg clioquinol, but a 65% decrease in Abeta in mice treated with 20 mg/kg. (PBT2 and clioquinol have similar molecular weights.) Additional findings indicating physiologic activity were seen at 30 mg/kg, including decreased serum Abeta and increased copper and zinc levels in the soluble cerebral fractions. Cherny 2001 writes:
We performed a pilot study of CQ treatment in the APP2576 Tg mice. We first compared the effects of CQ...on a cohort of 12-month-old APP2576 mice. The drugs were delivered by gavage daily for 12 weeks. The animals were sacrificed and brain Abeta levels were appraised. There was a mean decrease in the pellet fraction of cerebral homogenates from the animals treated with CQ 2 mg/kg/d group that did not reach statistical significance compared to sham-treated controls. However, there was a significant 65% decrease in the levels of sedimentable Abeta in the mice treated with CQ 20 mg/kg/d. Intriguingly, two animals in the CQ 20 mg/kg/d treatment group were found to have no measurable Abeta in the brain pellet fractions and no detectable amyloid pathology in their neocortex or cortical blood vessels.
In Crouch 2011, maximum phosphorylation of glycogen synthase kinase 3 occurred at 2.5 micromolar in cell culture. Serial dilutions in the range of 0-2.5 micromolar were not performed. Experiments evaluating the PBT2-mediated translocation of extracellular zinc into the intracellular space were performed at 10 micromolar. Studies demonstrating increased dendritic spine development (Adlard 2011) were performed at 0.15 micromolar in cell culture, but at oral doses of 30 mg/kg in intact animals. Follow-up studies showing that PBT2 protects wild-type mice from age-related cognitive decline (Adlard 2013) were also performed at 30 mg/kg.
In the key in vivo experiments (Adlard 2008), PBT2 at 30 mg/kg significantly reduced soluble by 30%, and insoluble Abeta by 37%, and reduced plaque burden by 80% with a p-value of 0.0008. PBT2 reduced the amount of plaque-bound zinc, reduced the inhibition of long-term potentiation by Abeta, improved 24-hour memory, improved learning, and improved performance in the Morris water maze. In this study, pharmacokinetics with measurement of plasma and brain concentrations of PBT2 were performed. PBT2 concentrations in brain interstitial fluid measured by in vivo microdialysis peaked 4 hours after 30 mg/kg gavage at about 25 nanomolar, with a half-life of roughly 4-6 hours. In wild-type animals, PBT2 concentrations were 796 +/- 514 ng/g (2.94 micromolar) in brain and 214 +/- 98 ng/ml (0.79 micromolar) in plasma 2 hours after oral administration. In transgenic Tg2576 mice, PBT2 concentrations were 318 +/- 75 ng/g (1.17 micromolar) in brain and 85 +/- 15 ng/ml (0.31 micromolar) in plasma 2 hours after oral administration. Concentrations measured in brain and plasma are much higher than the values recorded by microdialysis because of the high protein binding of PBT2 (about 95.1%). Thus, free PBT2 not bound to protein at 30 mg/kg oral dosing should be in the range of 57-144 nanomolar, which is in reasonable agreement with the in vivo microdialysis data. Clearly, the numbers are not exact.
Prana's dosing is based on the FDA's guidance for mouse-to-human equivalent dosing found here. The guidelines recommend reducing the dose per unit weight by a factor of 12.3 to convert from mouse dosing to human dosing. Assuming that 80 kg is a realistic weight for the average early Alzheimer's patient, the approximate starting dose for Prana should have been in the vicinity of 3.25 mg/kg, or 260 mg, which is very nearly the value used in the Phase 2 trials. The guidelines are very crude, however, and it would be naïve to think that the biodistribution for all drugs is going to be equivalent. 250 mg is a totally reasonable starting point. However, in light of the nearly-negative results from the Phase 2a Alzheimer's trial and the Reach2HD trial despite starkly positive results in animals, it is rational to hypothesize that the dosing is too low. Presumably, Prana has saved the CSF samples from the IMAGINE trial and will have the ability to measure PBT2 levels (and/or PBT2-Zn2+ complexes) on a post-hoc basis.
We can make some back-of-the-envelope calculations to estimate the CNS concentration of PBT2 in humans based on 250 mg dosing and an 80-kg patient with slightly above-average adipose tissue compared to healthy 20-year-olds. Assuming that 50% of total body weight is water and that non-complexed PBT2 will be distributed evenly, the expected plasma concentration would be 23 micromolar. Sheng 1979 reports that mice have a total body water of about 60-65% of their weight, so an initial estimate of PBT2 concentration after a 30 mg/kg dose would yield a concentration of 170-184 micromolar. Obviously, these concentrations were not achieved in vivo, possibly due to first-pass metabolism in the liver. If 30 mg/kg dosing in mice yielded brain concentrations of 1.17-2.94 micromolar, including the protein-bound fraction, it is plausible that dosing humans at 3.1 mg/kg yields brain concentrations that are closer to 0.12-0.30 micromolar, including the protein-bound fraction and free PBT2 concentrations of only 6-15 nanomolar. Low nanomolar concentrations in this range have not been observed to be effective in any preclinical biochemical, cellular, or in vivo model. It is not safe to assume that glycogen synthase kinase 3 phosphorylation and Abeta clearance will occur in physiologically meaningful quantities if the brain concentration in the human central nervous system is much lower than concentrations used in the preclinical experiments.
Because of the theoretical risk of permanent neurological injury associated with dramatically higher dosing, it would be preferable to study high-dose PBT2 in patients who have nothing to lose, such as those with more advanced Huntington's or Alzheimer's disease. The ethical issues surrounding high-dose experimentation on mentally impaired people are not straightforward, but I am confident that patients with these diagnoses are sufficiently desperate that patients or their family members will be more than willing to take large risks for the possibility of a successful treatment. At the very least, the neurotoxicology data from clioquinol in baboons suggest that there is a large range of dose increases that could be tolerated before reaching the danger zone.
Given the available data, I expect a mixed result from the IMAGINE trial. There may be some minimal cognitive benefit, but it will be apparent only with data contortions - perhaps a peculiar subgroup analysis, or a semi-acceptable statistical test. We are quite likely to see a decrease in CSF Abeta 1-42 levels, as in the Phase 2a data, but the magnitude of the decrease will probably be small. A small, borderline-significant decrease in amyloid on PiB PET scans is likely. I predict that there will be no difference in brain volume by MRI and probably no difference in fluorodeoxyglucose PET imaging. I consider it very unlikely that IMAGINE will have a powerful, incontrovertible result that will suffice for drug registration, and moderately improbable that the data will be strong enough to convince Prana's many doubters.
The best move for Prana after the results will be to measure the PBT2 levels in the CSF samples from the IMAGINE trial. Alternatively, subgroup analysis suggesting better efficacy in underweight or hypoalbuminemic patients could suggest that underdosing is taking place. If the levels of free PBT2 in the CSF are low (as I expect them to be), Prana should perform a short-term dosing trial in which the company identifies a higher safe dose. Such a trial could be performed in a few months at minimal cost. It need not be large, nor would such a trial need to include expensive PET scans. However, CSF sampling, possibly with tritiated or carbon-14 labeled drug, would be advisable to prove that adequate drug concentrations are being obtained in the CNS. It may also be helpful to study patients with more advanced disease than have been treated with PBT2 thus far. As soon as higher safe dosages are identified, the company could move into a definitive, pivotal Phase 3 trial with a reasonably high likelihood of success.
Authors suggesting that a company is overpriced often become targets of unhappy shareholders. I wish to emphasize that I am not "against" Prana and that I believe its hypothesis to be fundamentally correct. As a physician and decent human being, I desperately want Prana to find a cure for Alzheimer's disease, Huntington's disease, and other neurodegenerative disorders. I also believe that Prana's scientific staff are among the best in the world in Alzheimer's research. However, given the weak results reported hitherto, it would be in patients' (and investors'!) best interests if Prana revisited the dosing strategy and considered increasing the dose to levels more comparable to the animal models in which PBT2 was an incontrovertible success.
Additional disclosure: This is an exceptionally volatile and risky stock with short-term movements of +/- 40% being common. I may initiate a long or short position in the stock (possibly through options) at any time. If you find any errors in this article, please have the kindness to call them to my attention. Rely on your own due diligence for all of your investment decisions.