Curing Alzheimer's Disease

Introduction – the scale of the problem

Dementia is the persistent loss of previously acquired intellectual functions without impairment of consciousness (World Health Organisation). Although the prevalence of dementia increases with age, it is a disease, and not part of the normal ageing process. 800,000 people are currently diagnosed with dementia in the UK – this equates to a prevalence of roughly 1% for people aged 60-64 years, roughly doubling with every extra 5 years of age, reaching a prevalence of 40% in people aged 85-89 years (Alzheimer's Association). By 2021, it is predicted that over a million people in the UK will have dementia; in 2012, the estimated financial cost was already £23 billion.

The disturbing statistics and predicted trends with our ageing population has prompted the Health Secretary and Prime Minister to recently reiterate their commitment to finding a cure for dementia by 2025 (BBC). They believe this can be achieved through increased funding for research and offering the pharmaceutical industry extra incentives, such as prolonged patents for new drugs.

There are many causes of dementia, with Alzheimer’s disease (AD) being the most common – it is thought to represent 60-70% of cases. At older ages it often becomes more complicated as a combination of pathologies is present in around 50% of cases, typically AD mixed with vascular dementia, or AD mixed with Lewy-body dementia. What is irrefutable, however, is that a cure for AD remains the biggest and most important challenge in our goal to cure dementia.  Fortunately, much progress is being made. Two recent articles in Lancet Neurology Yan et al (Yan, 2014) and Dubois et al (Dubois, 2014) make this clear, and form the basis for this article.

The pathogenesis of Alzheimer’s disease

Potential therapeutic strategies arose from our increasing understanding of AD pathology. It has long been recognised that the brains of patients with AD are not only atrophic, but contain the specific microscopic features of amyloid plaques and neurofibrillary tangles:

• Amyloid plaques – extracellular aggregates of an abnormally fold protein, Aβ amyloid. This is surrounded by dystrophic neuritis and activated glial cells
• Neurofibrillary tangles – intracellular aggregates of the abnormally phosphorylated protein tau, surrounding the nucleus of neurones. Such aggregates persist after the death of such neurones

A number of years of vigorous scientific debate centred around which of these two features, the Aβ amyloid or the abnormal tau, was the initial causal abnormality in AD. Although it appeared that the amount of tau correlated more strongly with the severity of the disease (Braak, 1991), the discovery of genetically inherited forms of AD, where affected family members almost always developed the disease at a young age, provided strong evidence in support of Aβ (Rogaev, 1995) (Levy-Lahad, 1995). An understanding of how Aβ is formed is required before introducing these cases.

Amyloid Precursor Protein and the formation of Aβ

Amyloid precursor protein (APP) is a transmembrane protein that is subjected to proteolysis by multiple potential enzymes. Three of these are key to the understanding of AD pathology:

• β-secretase (aka β-site amyloid precursor protein cleaving enzyme 1, BACE) – cuts APP so to release the distal fragment, known as soluble APP β (sAPPβ)
• γ-secretase – acts after BACE, cutting APP centrally within its transmembrane region, releasing the Aβ. Depending exactly where it cuts, it either forms Aβ with a length of 40 or 42 amino acids (Aβ_1-40 or Aβ_1-42); Aβ_1-42 is known to be more toxic.
• α-secretase – cuts APP slightly more centrally than BACE does, within the Aβ segment, therefore preventing the formation of Aβ protein. The fragment this releases in known as soluble APP α (sAPPα)

                                                                             

Familial cases of Alzheimer’s disease

More than 200 autosomal dominant mutations are now known to cause AD (Yan, 2014), all of which lead to either an increase in the total production of Aβ, or an increase in the proportion of the more toxic Aβ_1-42 form.

• Mutations in APP result in increased hydrolysis by BACE
• Mutation in presenilin, the catalytic subunit of γ-secretase, increase the formation of Aβ_1-42
• Mutations in ADAM10, the main α-secretase in neurones, decreases its activity, hence more Aβ is instead produced via BACE
• Duplications of APP, either in isolation, or as part of trisomy 21, result in early onset AD (patient’s with Down’s syndrome typically develop the disease around 40 years of age)

Furthermore, mutations in APP that reduce the ability of BACE to cleave it are protective again AD, and mutations in tau proteins have been shown to be associated with other forms of dementia, such as fronto-temporal dementia, but do not seem to increase the risk of AD (Hutton, 1998).

Conclusions – known causes and potential targets

As a result of this genetic evidence, there is a strong case for the primary cause of AD to be the formation of Aβ, particularly Aβ_1-42, with the associated tau pathology being a downstream consequence of this. Together, this is known as the amyloid cascade hypothesis (Higgins, 1992). Furthermore, the existence of protective mutations that act via a reduction in BACE activity makes it a strong candidate drug target. Other strategies based upon the amyloid cascade hypothesis involve using immunotherapies to target Aβ_1-42. Phase 3 trials of both solanezumab (Doody, 2014) and bapineuzumab (Salloway, 2014) in patients with mild-moderate AD have, however, failed to show differences in cognitive outcomes. As will be discussed later, this may be due to the timing of treatment, rather than lack of potency.

Mouse models

Transgenic mice designed to over-express APP develop amyloid plaques and memory impairment, apparently mimicking AD in humans (Saito, 2014). In order to model the potential for drugs that inhibit BACE, further transgenic mice with over-expression of APP but no BACE genes were produced. Such knock-out mice do not develop amyloid plaques or memory impairment. They do, however, experience side effects – reduction in the myelination of neurones, axonal guidance defects and neurogenesis problems, to name as few. The exciting prospects such models suggests for treatments in humans must therefore be interpreted with caution – drugs inhibiting BACE may reduce the amount of Aβ produced and so potentially help in AD, but they may also cause problems. We known that BACE enzymes do not only act upon APP, but also on many other proteins found on the surface of neurones (Yan, 2014). If you prevent BACE from producing Aβ, you also therefore prevent it from producing a number of other protein. Many of these are thought to be involved in the communication within a neuron and between nearby neurones, and hence the side effects that have been seen in mice.

The development of BACE inhibitors

Following the results of the BACE knock-out mice, a number of inhibitors went into development, hoping to produce a drug that could be taken orally and result in the same prevention of Aβ amyloid production. Initial difficulties with oral bioavailability and blood-brain barrier penetration were eventually overcome, leading to a number of such inhibitors entering clinical trials in humans. The main outcome measures were safety and tolerability of the BACE inhibitors, as well as efficacy as measured via CSF Aβ concentrations.

Drug	MK-8931	LY2886721	E2609
Trial	Phase 1b RCT vs placebo, n=32, mild-mod AD patients, 7 days treatment	Phase 1 RCT vs placebo, n=47, healthy volunteers, 14 days	2x phase 1 RCT vs placebo, n= 73 and n=53
Maximum CSF Aβ reduction	84%	74%	85%
Major side effects	Nil	Nil	Nil

What level of BACE inhibition is required to treat Alzheimer’s disease?

The results of the initial phase 1 trials of BACE inhibitors are very promising, but a number of questions remain unanswered. The absence of major side effects in such short trials is not unexpected; the concerns regarding side effects that arose from the mouse knock-outs, such as problems with myelination, neurogenesis and axon guidance, would be unlikely to manifest in humans over such short periods. One potential strategy to avoid such effects would be to use lower levels of inhibition. Homozygous knock-outs clearly represent 100% inhibition of BACE activity – an unachievable and potentially undesirable figure with oral inhibitors. Support for targeting lower levels of inhibition comes from two sources:

• The protective mutation in APP that reduces the ability of BACE to cleave it. Heterozygous humans have around a 20% overall reduction in cerebral Aβ production; this is enough to protect them against AD, but they also do not seem to suffer from any of the other potential problems
• Heterozygous BACE knock-outs in AD mouse models have a 50% reduction in BACE activity, resulting in around a 20% reduction in Aβ production. They do not develop amyloid pathology or memory deficits, and neither do they develop the side effects that homozygous knock out mice do.

It would therefore seem that if drugs could achieve 50% inhibition of BACE activity, the overall Aβ production might drop by around 20%; this would hopefully be enough to prevent AD pathology from forming without the patients developing any other side effects from the reduced BACE activity.

When should we be using BACE inhibitors to treat Alzhiemer’s disease?

In both the heterozygous human and mouse examples above, the reduction in BACE activity is clearly present from the moment of conception. There is a life-long reduction in Aβ production, and this leads to the protective effect. A difficult question therefore remains – when using BACE inhibitors to treat AD, at what stage would they need to be started? It is well known from studies of familial AD that the neuropathological changes begin long before the disease becomes symptomatic, and it is unlikely that inhibiting BACE would reverse the neuropathological changes once the amyloid plaques and neurofibrillary tangles have already formed. It therefore seems likely that in order to be effective, such drugs would need to be given long before symptoms develop; rather than treating AD, they should be seen as more of a preventative strategy.

At the heart of the question on when to use BACE inhibitors is the problem of our understanding of the exact role of Aβ in AD pathophysiology. It is thought the Aβ oligomers are more likely to mediate neuropathology than the larger aggregates (Shankar, 2008), but it is still the tau neurofibrillary tangles that correlate with neurodegeneration, not the amyloid plaques. We also know from familial fronto-temporal dementia that isolated tau pathology is sufficient to cause neurodegeneration and dementia. The amyloid cascade hypothesis therefore has room for various situations (Karran, 2011):

1. Aβ amyloid trigger – the production of Aβ triggers the formation of tau pathology, which is then self perpetuating
2. Aβ amyloid threshold – Aβ results in self-perpetuating production of tau once a sufficient threshold level of Aβ is produced
3. Aβ amyloid driver – the continuous production of Aβ is necessary in order to perpetuate tau pathology and neurodegeneration

The amyloid trigger and amyloid threshold hypotheses would require earlier use of BACE inhibitors, preventing the triggering or threshold levels of Aβ from arising. The amyloid driver hypothesis would be more hopeful, suggesting that even in people with existing AD pathology, reductions in Aβ production would lead to a plateau in neurodegeneration.

The above questions are only going to be answered a combination of more basic research furthering our knowledge of the interaction between Aβ, tau and neurodegeneration, and clinical trial results using agents that reduce Aβ production, such as BACE inhibitors. We can already be fairly sure, however, that when it comes to reducing Aβ production, earlier is certainly going to be better.

Current phase 3 trials using the agents mention above are recruiting patients with mild-moderate AD or mild cognitive impairment (10-15% of which convert to full AD every year). Their primary outcome measures will involve cognitive assessments, but also CSF biomarkers and PET amyloid imaging; the data produced will, therefore, undoubtedly greatly advance our understanding of the potential for BACE inhibitors and how best to use them. It would seem reasonable to predict, however, that they will not “cure” many of the patients – cognitive stabilisation is probably the best that can be hoped for (via the amyloid driver hypothesis). A potential cure would require earlier treatment – it is thought that this may have contributed to the failure of the phase 3 trials using immunotherapies in mild-moderate AD patients (Doody, 2014) (Salloway, 2014).

Pre-symptomatic diagnosis allowing disease prevention

The analogy of our advances in prevention of heart disease is put forward by Yan et al (Yan, 2014). When someone has a heart attack, they will be put on statins (amongst other drugs). These will significantly reduce their cholesterol, hence reducing their risk of having further heart attacks. This might prevent their heart from getting too much worse (or at least slow the rate of decline), but they are not going to reverse the damage that has already happened. The good thing about heart attacks, however, is that you can, with reasonable accuracy, predict the chance that someone is going to have a heart attack. Various tools, such as the widely used Q-risk, can be used to predict an individual’s risk based upon factors such as their age, sex, ethnicity, smoking status, diabetic status, cholesterol levels, blood pressure and body mass index. When their risk is sufficiently high, drugs such as statins can be started years in advance of any heart attack they might have, greatly reducing their risk of it ever happening.

Once AD becomes symptomatic, the best that we can realistically hope for is that BACE inhibitors will be able to slow the rate of decline, but not reverse the damage. What is therefore needed is the equivalent of a Q-risk for AD, allowing us to predict who is going to be affected years in advance of their first symptoms arising. This will allow prevention of disease by avoiding the amyloid trigger or amyloid threshold. Progress towards this is being made, as outlined by the publication of the International Working Groups second diagnostic criteria of AD (Dubois, 2014).

Biomarkers in the diagnosis of AD

The diagnosis of AD has previously relied upon clinical features. The original 1984 criteria relied upon clinical features, only allowing the diagnosis to be “probable AD” upon confirmation with post-mortum histology (McKhann, 1984). Since then, however, our knowledge of biomarkers has greatly improved, allowing them to be incorporated into diagnostic criteria.

CSF biomarkers

CSF levels of Aβ_1-42 inversely correlate with cerebral amyloid burden – the more amyloid plaques that are present, the lower the CSF Aβ_1-42 levels. When correlated with post-mortum histology, a low CSF Aβ_1-42  had a 96.4% sensitivity for the diagnosis of AD. Low levels are also seen in some other dementia, however, such as Lewy-body dementia and vascular dementia, limiting the specificity. This, however, can be improved by instead using the Aβ_1-42 : Aβ_1-40 ratio – a low ratio is more specific to AD.

CSF tau levels have also been shown to be predictive, with higher levels being able to distinguish AD from other non-AD dementias. Both P-tau and T-tau can be used: T-tau is thought to be a direct reflection of neuronal degeneration, where as P-tau reflects the total tau burden.

Overall, a combination of tau and Aβ appears to be the best strategy for interpretation of CSF biomarkers; the T-tau to Aβ_1-42 ratio had a 96% specificity for distinguishing between FTD and AD, and the ADNI, DESCRIPA and SBP projects have all confirmed the utility of the combined biomarkers. Perhaps most interestingly of all, the DIAN project, using a cohort of patients with autosomal dominant AD associated mutations, has demonstrated that significant predictive changes in such CSF biomarkers are present 10-20 years prior to the onset of the first symptoms (Bateman, 2012). This opens the door for effective, early preventative treatment with β-secretase inhibitors.

Bateman. (2012). Clinical and Biomarker Changes in Dominantly Inherited Alzheimer’s Disease - reductions in CSF Aβ together with increases in CSF tau PET amyloid deposition occur up to 20 years before diagnosis of AD (defined as year = 0)

Imaging biomarkers

PET imaging with various ligands has demonstrated the ability to accurately quantify the cerebral load of Aβ amyloid. This correlated well with post-mortum studies (Clark, 2011).

Clark. (2011). Use of florbetapir-PET for imaging beta-amyloid pathology - correlation of amyloid PET imaging with post-mortem histology confirming the strong predictive value of PET

When the CSF biomarkers are compared with PET amyloid imaging, a strong association is also seen – a reduction in soluble Aβ and tau in the CSF correlates with those who have higher cerebral amyloid burdens (Roe, 2013) (Fagan, 2011). This is a very positive finding in terms of validating the biomarkers – it would seem to support the notion that a reduction in CSF amyloid occurs because more of it is being deposited in the brain parenchyme; as this happens, with the initiation of neurodegeneration, levels of tau begin to rise.

Fagan. (2011). Receiver operating curves comparing the ability of various CSF markers to predict whether amyloid PET imaging was positive or negative in a cohort of patients who were either cognitively normal or with mild cognitive impairment; the p-tau/Aβ1-42 ratio had the greatest accuracy.

A target for clinical trials – “asymptomatic at risk for Alzheimer’s disease”

In light of the above evidence for the value of biomarkers in diagnosis of AD long before symptoms arise, the IWG-2 criteria now include the following definition (Dubois, 2014):

Asymptomatic at risk AD (A plus B)

a.     Absence of specific clinical phenotype

                                               i.     Absence of amnestic syndrome of the hippocampal type

                                              ii.     Absence of clinical phenotype of atypical AD

b.     In-vivo evidence of Alzheimer’s pathology (one of the following)

                                               i.     Decreased Aβ_1-42 together with increase T-tau or P-tau in CSF

                                              ii.     Increased retention of fibrillar amyloid PET

The acceptance of this category is a landmark event in AD research, representing the extent to which the field of biomarkers has advanced in recent years. It clearly states that either a combination of CSF Aβ and tau or PET imaging is sufficient to determine whether an asymptomatic patient should be placed in the asymptomatic at risk category.

Conclusions

The papers by Yan et al and Dubios et al summaries two areas where significant developments have recently been made in the field of AD. The true significance of these advances, however, is only revealed when they are taken together – the development of new drugs that could potential prevent AD from developing if started early enough, and the development of new biomarkers that allow presymptomatic diagnosis.

Any clinical trials to test such combined potential would still, however, be demanding to complete. With the potential to diagnose patients 10-15 years before symptoms develop comes the commitment to continue clinical trials for 10-15 years after randomised treatments are started in such individuals.  Perhaps this is where the promises of the Health secretary and the Prime Minister may come in – increases in funding and government-pharmaceutical partnerships may be required in order to facilitate such studies.

Examples of such partnerships already exist in the trials of immunotherapies in AD. The Anti-Amyloid Treatment in Asymptomatic Alzheimer’s study (A_4, NCT02008357) is recruiting asymptomatic individuals who are at risk of developing AD as determined by a PET amyloid scan. Funding is being provided by a partnership between the National Institute on Ageing and Eli Lilly. Those meeting the criteria will the randomised to receive solanezumab, an anti- Aβ_1-42 antibody, or placebo. Other similar immunotherapy studies recruiting asymptomatic patients are also on-going (Dominantly Inherited Alzheimer Network Trial (DIAN-TU; ClinicalTrials.gov number, NCT01760005), Alzheimer's Prevention Initiative (API; NCT01998841)). One hopes that such trials can produce better outcomes than the results in early AD patients. If this occurs, and similar trials can be initiated for BACE inhibitors, perhaps by 2025, as targeted by the Prime Minister, we will not only have one cure for Alzheimer’s disease, but two.

Bibliography

Alzheimer. (1907). About a peculiar disease of the cerebral cortex. Centralblatt für Nervenheilkunde Psychiatrie (30), 177-179.

Alzheimer's Association. (n.d.). Alzheimer's Disease Fact Sheet. From http://www.alzheimers.org.uk/site/scripts/documents_info.php?documentID=341

Bateman. (2012). Clinical and Biomarker Changes in Dominantly Inherited Alzheimer’s Disease . New England Journal of Medicine .

BBC. (n.d.). From http://www.bbc.co.uk/news/health-27912473.

Braak, B. a. (1991). Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol (82), 239-259.

Clark. (2011). Use of florbetapir-PET for imaging beta-amyloid pathology. . JAMA (305), 275-283.

Doody. (2014). Phase 3 Trials of Solanezumab for Mild-to-Moderate Alzheimer's Disease . New England Journal of Medicine (370), 311-321.

Dubois. (2014). Advancing research diagnostic criteria for Alzheimer's disease: the IWG-2 criteria. Lancet Neurology (13), 614-629.

Fagan. (2011). Comparison of analytical platforms for cerebrospinal fluid measures of Aβ1-42, total tau and p-tau181 for identifying Alzheimer’s disease amyloid plaque pathology . Arch Neurol .

Goate. (1991). Segregation of a missense mutation in the amyloid precursor protein gene with familial Alzheimer’s disease. Nature (349), 704-709.

Higgins, H. a. (1992). Alzheimer’s disease: the amyloid cascade hypothesis. Science (256), 184-185.

Hutton. (1998). Association of missense and 5′-splice-site mutations in tau with the inherited dementia FTDP-17. Nature (393), 702-705.

Karran. (2011). Thee amyloid cascade hypothesis for Alzheimer’s disease: an appraisal for the development of therapeutics . Nature Reviews Drug Discovery , 10.

Levy-Lahad. (1995). Candidate gene for the chromosome 1 familial Alzheimer’s disease locus. Science (269), 973-997.

McKhann. (1984). Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology .

Roe. (2013). Amyloid imaging and CSF biomarkers in predicting cognitive impairment up to 7.5 years later . Neurology .

Rogaev. (1995). Familial Alzheimer’s disease in 41 in kindreds with missense mutations in a gene on chromosome 1 related to the Alzheimer’s disease type 3 gene. Nature (376), 775-778.
Saito et al. (2014). Single APP knock-in mouse models of Alzheimer's disease. Nat. Neurosci 17(5):661-663

Salloway. (2014). Two Phase 3 Trials of Bapineuzumab in Mild-to-Moderate Alzheimer's Disease. New England Journal of Medicine (370), 322-333.

Shankar. (2008). Amyloid-β protein dimers isolated directly from Alzheimer’s brains impair synaptic plasticity and memory. Nature Med. (14), 837-842.

World Health Organisation. (n.d.). Alzheimer's Disease Fact Sheet. From http://www.who.int/mediacentre/factsheets/fs362/en/

Yan. (2014). Targeting the beta-secretase BACE1 for Alzheimer's disease therapy. Lancet Neurology (13), 319-329.