An exciting new study on Alzheimer’s disease (AD) was published in the 2 August issue of Nature. The study was carried out by researchers at deCode Genetics (Reykjavik Iceland) and their collaborators at Genentech and several academic institutions. A News and Views article by leading AD researcher Bart De Strooper and genomics researcher Thierry Voet (both at KU Leuven, Leuven, Belgium) analyzes this study and its implications.
Amyloid plaques are a central feature of AD. They largely consist of amyloid-β (Aβ) peptides. Aβ peptides are formed via sequential proteolytic processing of the amyloid precursor protein (APP), catalyzed by two aspartyl protease enzymes–β-secretase and γ-secretase. The β-site APP cleaving enzyme 1 (BACE1) cleaves APP predominantly at a unique site. However, γ-secretase cleaves the resulting carboxy-terminal fragment at several sites, with preference for positions 40 and 42. This leads to formation of amyloid-β1–40 (Aβ40) and Aβ1–42 (Aβ42) peptides. APP processing to yield Aβ peptides is illustrated by the figure at the top of this article.
By studying rare, familial cases of early-onset AD, human geneticists have identified three disease genes in these conditions— genes for APP, and for two presenilins, PS1 and PS2. The presenilins are components of γ-secretase, which exists as an intramembrane protease complex. Mainly because of these genetic studies, as well as studies in animal models and postmortem studies of AD brains, the majority of AD researchers have focused on the APP processing pathway and/or on aggregation of Aβ to form plaques as intervention points for therapeutic strategies. The hypothesis that this is the central AD disease pathway is called the “amyloid hypothesis”.
Up until the publication of the new deCode report, of the 30-odd coding mutations in APP that have been found, around 25 are pathogenic, usually resulting in autosomal dominant early-onset Alzheimer’s disease. Coding mutations at or near the β- or γ-proteolytic sites have appeared to result in overproduction of either total Aβ or a shift in the Aβ40:Aβ42 ratio towards formation of Aβ42, which is the more toxic of the two Aβ peptide. Until now, mutations in APP have not been implicated in the common, late-onset form of Alzheimer’s disease.
In the new deCode study, the researchers studied coding variants in APP in a set of whole-genome sequence data from 1,795 Icelanders. They identified a single nucleotide polymorphism (SNP), designated as rs63750847. The A allele of this SNP (rs63750847-A) results in an alanine to threonine substitution at position 673 in APP (A673T). The A673T mutation was found to be significantly more common in the elderly (age 85-100) control group (i.e., those without AD) than in the AD group. The researchers therefore concluded that the mutation is protective against AD.
The researchers also found that in a cohort of individuals over 80, those who were heterozygous for the A673T mutation performed better in a test of mental capacity than did control subjects. The authors concluded that the A673T mutation not only protects against AD, but also against the mild cognitive decline that is normally associated with old age.
In cellular studies (i.e., studies in cultured cells transfected with genes coding for wild type or mutant APP) and in biochemical studies, the researchers found that APP carrying the A673T mutation undergoes about 40% less cleavage by BACE1 than does wild-type APP, resulting in 40% less production of both Aβ40 and Aβ42.
The researchers conclude that the strong protective effect of the A673T mutation against AD provides proof of principle for the hypothesis that reducing the β-cleavage of APP (e.g., by use of BACE1 inhibitors, such as those being developed by some pharmaceutical companies) may protect against the disease. (However, success in developing BACE1 inhibitors has been elusive.) Moreover, since the A673T allele also protects against cognitive decline in elderly individuals who do not have AD, AD and age-related mild cognitive decline may be mediated through the same or similar mechanisms.
Despite this compelling genetic finding, amyloid pathway-targeting drugs have not shown efficacy in Phase 3 trials
In our January 26, 2010 blog article, we discussed Phase 2 clinical trials of bapineuzumab, a monoclonal antibody (MAb) drug that is specific for Aβ, in mild to moderate AD. In that article, we referred to the drug as “Elan/Wyeth’s bapineuzumab”, after the original developers of the drug. As the result of mergers and acquisitions, the drug is now referred to as “Pfizer/Janssen’s bapineuzumab”. Many commentators call it “bapi” for short.
As we discussed in that article, the overall result of the Phase 2 trial was that there was no difference in cognitive function between patients in the bapi-treated and the placebo groups. However, the study did not have sufficient statistical power to exclude the possibility that there was such a difference. Retrospective analysis of the data from the trial suggested that bapi-treated patients who were not carriers of the apolipoprotein E epsilon4 allele (ApoE4) showed improved cognitive function as compared to placebo treatment. Given that this conclusion was reached via retrospective analysis, the idea that the bapi was efficacious in ApoE4 noncarriers was only a hypothesis, which would require prospective clinical trials to confirm. Janssen and Pfizer had been conducted large Phase 3 trials of bapi, which they prospectively segregated into ApoE4 carrier and noncarrier groups in order to test this hypothesis.
As of the past several weeks, the results of these Phase 3 trials have come in. On July 23rd, 2012, Pfizer announced the top-line results of an 18-month Janssen-led Phase 3 study of intravenous bapi in approximately 1,100 patients with mild to moderate Alzheimer’s disease who carry at least one ApoE4 allele. The drug failed to meet its co-primary endpoints (change in cognitive and functional performance compared to placebo) in that study. On August 6, 2012, Pfizer announced the top-line results of the corresponding Phase 3 study of intravenous bapi in patients with mild-to-moderate Alzheimer’s disease who do not carry the ApoE4 genotype. Once again, the co-primary clinical endpoints were not met. Based on these results, the companies decided to discontinue all other intravenous bapi studies in patients with mild-to-moderate Alzheimer’s disease.
The bapi development program continues a history of amyloid pathway-targeting drugs that were taken into Phase 3 trials despite Phase 2 results that showed no statistically significant efficacy. For example, we cited the cases of Myriad Pharmaceuticals’ Flurizan (tarenflurbil) and Neurochem’s (now Bellus Health) Alzhemed (3-amino-1-propanesulfonic acid) in our January 26, 2010 blog article.
Leading industry commentator Matthew Herper of Forbes referred to the failure of bapi as “the latest piece of evidence of the drug industry’s strange gambling problem.” Johnson & Johnson (the parent company of Janssen) spent more than $1 billion to invest in Elan and get one-quarter of bapi, and Wyeth (later Pfizer) and Elan put the drug into Phase 3, despite the Phase 2 failure of bapi.
The temptation for pharmaceutical companies to take a chance on an AD drug such as bapi, Flurizan, and Alzhemed is driven by the complete lack of disease-modifying AD drugs, and the thinking that even a not-very-effective drug that receives FDA approval might generate billions of dollars in annual sales. In the case of bapi there was also that tantalizing suggestion that bapi might show efficacy in the subset of patients who lacked ApoE4.
In an August 16, 2012 article in Forbes, Dr. John LaMattina (the former President of Pfizer Global R&D) engages in informed speculation as to why bapi was moved into Phase 3. Dr. LaMattina (in contrast to critics like Mr. Herper, who discounted the ApoE4 retrospective analysis as “data-dredging” that was “likely to be due to chance”) referred to the efficacy signal of the Phase 2 trials as “mixed” due to the ApoE4 analysis. He stated that such “mixed results” present an “agonizing” dilemma for a pharmaceutical company.
In deciding whether to go forward Phase 3 trials of bapi, Dr. LaMattina further speculates that the decision might have been influenced by stakeholders such as AD patient advocates, and scientists who strongly believed in the science behind bapi, especially the amyloid hypothesis. Moreover, bapi had been shown to be relatively safe. In addition, dropping bapi would have caused public relations damage. Dr. LaMattina concludes, based on this analysis, “…this was a situation where these companies were in possession of a relatively safe drug, with a modest chance of success in being efficacious in what may be the biggest scourge that society will face. How can you not make this investment?” He reminds us that pharmaceutical R&D “is a high risk, high reward business”.
Nevertheless, bapi joined Flurizan and Alzhemed on the list of high-profile amyloid-pathway failures. Now a Phase 3 trial of Lilly’s solanezumab, another MAb drug that targets Aβ, is nearing completion, with the results expected in September. Published Phase 2 results were designed to test safety, not efficacy, and 12 weeks of drug treatment gave no change in cognitive function. Although the results of the Phase 3 trial will not be known until they are reported, analysts expect the drug to fail because of its similarity to bapi.
Why don’t amyloid pathway-targeting drugs show efficacy in clinical trials, despite the compelling genetic evidence for the amyloid hypothesis?
The almost standard answer to that question given by scientists and clinicians who support the amyloid hypothesis is that we have been testing the drugs too late in the course of AD progression, after the damage to the brain has become irreversible. Roche/Genentech is testing this idea in its clinical trials of its drug candidate crenezumab (licensed from AC Immune), which is yet another MAb drug that targets Aβ. In a 5-year Phase 2a clinical trial, Genentech is testing intravenous crenezumab in 300 cognitively healthy individuals from a large Colombian kindred who harbor the Glu280Ala (codon 280 Glu to Ala substitution) PS1 mutation. This mutation causes dominant early−onset familial AD, and is associated with increased levels of Aβ42 in plasma, skin fibroblasts, and the brain. Family members with this mutation begin showing cognitive impairment around age 45, and full dementia around age 51.
Genentech is conducting this trial in collaboration with the Banner Alzheimer’s Institute and the National Institutes of Health. The company says that this trial is the first-ever AD prevention study in cognitively healthy individuals. Genentech further says that the trial may help to determine if the amyloid hypothesis is correct–more specifically, it may help to determine if a drug that works by depleting amyloid plaques can be effective in preventing and/or treating AD.
Moreover, Genentech states that there is significant unmet medical need within this Colombian population. This large extended family may have as many as 5,000 living members, and no other population in the world offers a sufficiently large number of mutation carriers close to the age of potential disease onset for a study to determine whether a prevention treatment may work. This effort by Genentech thus represents an application of a rare disease strategy to AD.
It is also possible, however that drugs that work by lowering levels of Aβ will not be efficacious in treating AD, even if administered early in the disease process. This may be true despite the findings of the new genetic study by the deCode Genetics group. For example, in their Nature News and Views article, Drs. De Strooper and Voet remind us that if the A673T mutation indeed works via lowering of Aβ levels, it works via lifelong lowering of Aβ, not lowering of Aβ in patients who already have AD, as in all clinical trials so far of anti-Aβ antibodies. (Even Genentech’s Colombian trial may involve lowering of Aβ levels relatively late in the course of exposure of patients to a disease process that will result in AD.)
Moreover, as these authors speculate on the basis of work on another mutation at the same site in the APP protein, it is possible that the protective effect of the A673T mutation may be due to changing the aggregation properties of Aβ peptides, resulting in a less-toxic form of Aβ. If true, this would mean that the protective effect of the A673T mutation is due to qualitative, rather than quantitative changes in Aβ. In that case, the finding of protection from AD by the A673T mutation might not be as predictive of the efficacy of such Aβ-lowering treatments as the use of anti-Aβ MAb drugs as drug developers might like.
As the producers of this blog, and as consultants to the biotechnology and pharmaceutical industry, Haberman Associates would like to hear from you. If you are in a biotech or pharmaceutical company, and would like a 15-20-minute, no-obligation telephone discussion of issues raised by this or other blog articles, or of other issues that are important to your company, please click here. We also welcome your comments on this or any other article on this blog.