How Not to Study a Disease: The Story of Alzheimer's

How Not to Study a Disease; The Story of Alzheimer's. by Karl Herrup

In How Not to Study a Disease, neurobiologist Karl Herrup explains why the Alzheimer's discoveries of the 1990s didn't bear fruit and maps a direction for future research. Herrup describes the research, explains what's taking so long, and offers an approach for resetting future research.

Alzheimer’s disease (AD) has been notoriously hard to define. Herrup offers a working definition as a starting point: Alzheimer’s is a late life disease that, over the course of many years, destroys normal brain function in a progressive and irreversible fashion. It includes the progressive inability to form new memories, perform complex tasks, use language and reasoning, navigate an environment, and ability to make judgments. Personality changes such as depression, apathy, emotional outbursts, aggression, and agitation also follow. Usually, the sufferer lacks awareness of these changes.

Risk Factors: Women tend to get AD at about twice the rate as men, and prevalence also clusters geographically. However, both of these trends are largely explained by lifespan. Aging is the primary risk factor. Higher education is associated with lower AD risk, but it is not clear which direction the causality flows. Insulin resistance is also a major risk factor.

The form of Apolipoprotein E (ApoE) you carry is the main genetic risk factor for late onset AD. ApoE is a lipid carrying protein found in the brain as well as the blood. One form of ApoE4=5x elevated risk of AD, two forms=10x elevated risk of AD. ApoE4 is less efficient at carrying cholesterol than the other forms. ApoE carries amyloid as well. There’s also a link with myelin, which is mostly cholesterol, and as we age, the cells that make up myelin need larger volumes of cholesterol carried to them.

Early vs. Late AD: early onset tends to start before age 65, and to be more aggressive and run in families more. Late onset, or sporadic, accounts for about 95% of AD cases. However, most of the research on AD has been done on early onset. For example, transgenic mouse models are primarily early-onset AD models.

Genetics: In comparisons of genomes of people with sporadic AD and without, areas that differ are clustered in areas of the genome that regulate immune function, inflammation regulation, fat handling, vesicle trafficking (especially the specialized autophagosomes that deal with cellular garbage, efficiency degrades as you age).

Other biological factors: calcium regulation worsens with age, and these changes happen faster in AD, age-related myelin loss is sped up in AD (disappears first where it was added last), viral DNA and protein have been found in AD brains and cells in AD brains have molecular signs that they’ve been fighting a viral infection (amyloid may be antibacterial and trying to help?). Vessicles in AD brains contain proteases normally only found inside lysosomes, indicating a role for garbage disposal dysfunction in AD.

Plaques and Tangles: The distribution of plaques doesn’t correlate well with cognitive dysfunction in AD. Tangles correlate better. Plaques are composed of a segment of the larger amyloid protein that aggregates in Beta sheets (Amyloid-beta).

Microtubules are like the beams and girders of cells to give cells tensile strength and shape. When tau proteins bind to the microtubules, it makes them stiffer. When a phosphate group is added to tau by a kinase, it tends to come off the microtubule. In AD, the kinases are overactive, and add too many phosphates. Tau comes off microtubules in excess, which makes microtubules too floppy. Also, when tau is over phosphorylated, it tends to aggregate. So aggregated tau along with floppy microtubules makes tangle. Cell dies.

The two processes may be related. There’s an idea that the amyloid-beta peptide, through its regulation of calcium, causes tau to be over phosphorylated, which sets off tau tangles.

The drug discovery process: Fewer than 1 in 10 drugs that start phase I clinical trials end up with FDA approval. Basic science is usually conducted at academic institutes, and clinical trials are often run by drug companies because they are so expensive and risky. The suppression of non-amyloid research ended up creating a bevvy of inbred experts that the industry repeatedly called upon in the drug development process for AD. Even when initial amyloid clearing drugs failed to improve cognitive symptoms, new trials were started to treat earlier in the disease process, picking participants more selectively, drugs that targeted amyloid in a different way etc.

Beta- and gamma- secretase inhibitors were also tried, but they had huge off target effects because APP exists in almost every cell type in the body and those secretases cut it for various functions.

The heady amyloid years:

• A team at UCSD gets access to amyloid plaques (via blood vessels that feed the brain, which might not be exactly the same type of plaques, but people thought at the time that the plaques only happened in AD, so it seemed reasonable), and figure out the amino acid sequence of it

• Down syndrome brain was loaded with amyloid, and amino acid sequence showed it to be very similar to the AD plaques imaged by the UCSD team. Down syndrome is caused by an extra copy of chromosome 21, and the gene for the main plaque protein (amyloid-beta peptides) mapped to that chromosome!

• Precursor protein for amyloid-beta peptides is tracked down, looks like a cell surface receptor protein that sticks into the cell membrane. They called it amyloid precursor protein (APP).

• In the 1990s, geneticists found that in familial, early-onset AD, mutations appeared very near where the amyloid-beta peptide gene is found on chromosome 21, and led to increased production of APP. Two more mutations were identified, presenilin 1 (chromosome 14) and presenilin 2 (chromosome 1), which are both associated with gamma-secretase, an enzyme that cuts APP. The presenilin mutations cut amyloid-beta peptides out of APP in a way that made the amyloid-beta peptides more prone to aggregation (a longer 42 amino acid form vs a shorter 40 amino acid form)

• Human beings are the only species that get amyloid plaques and associated dementia, for reasons not yet clear, even though most species have the APP gene as well as the two secretases that cleave it into amyloid-beta. So transgenic mice were created to carry the human version of the APP gene with mutations present in familial early-onset AD. Elan pharmaceuticals developed a vaccine using artificially synthesized amyloid-beta to train the immune system to recognize and clear amyloid-beta, which worked super well in mice to prevent and actually clear already formed plaques. But in human trials, the vaccine caused life-threatening side effects. Later, we did find a way to clear plaques (passive immunization, but that didn’t help either).

Thus the amyloid cascade idea was formed. Because of the genetics of early-onset AD, plaques became the main focus of the field. The idea that amyloid-beta plaques caused AD was so strong that in future trials, amyloid-beta in CSF was often used as an outcome measure in the absence of cognitive measures. Alternative hypotheses, like that the beta- and gamma- secretases that cleave APP into amyloid-beta also regulate myelin thickness, and so it is possible that the presenilin mutations are linked to familiar AD not because of their role in amyloid-beta cleavage, but because of myelin, are relegated to the fringe.

Oh shoot, it might not be amyloid after all

• Plaque location doesn’t correlate well with cognitive dysfunction

• 25-30% of elderly people with normal cognition have a density of amyloid plaques in their brain that is similar to AD brains. Almost everyone has some, even kids. PiB staining in living brains using PET starting in 2002 started to make this clear. Having amyloid deposits in the brain increases risk of developing AD by 3-4 fold over the next few years, but it seems like our brains can function pretty well in the presence of a lot of plaques. Author doesn’t discuss cognitive reserve.

• 15% of people clinically diagnosed with AD don’t have significant plaque formation

• It turns out there’s a long delay between the appearance of plaques and the start of dementia, suggesting a third factor might be causing both. Alternatively, it’s possible the brain has enough redundant pathways to work around plaques up to a certain threshold of plaque density.

• Mice with human versions of mutated APP show plaque build up, but very few other health problems. They do normal species-typical behaviors, and only develop subtle cognitive deficits, if at all. We had trouble replicating the cognitive deficits in our own lab during my research. These mouse models often lack tau tangles, neuronal death, synapse loss, and other hallmark features. Yet they were used as the main model of AD.

• When mouse AD models do get cognitive deficits, they’re not very AD-like. Also, they don’t degenerate over time.

• PiB staining showed that early and late onset AD had pretty different patterns of plaque deposition in humans, not clear they’re actually the same disease since plaques are pretty non-specific pathology in the first place

All this actually shows the mouse and human data align well. Mice can have a lot of amyloid-beta plaques, and not have cognitive symptoms, just like humans.

The great inflations:

• Alois Alzheimer discovered a presenile dementia that resulted in severe memory impairment. Upon death, he found that brains of these people had aggregations of a peculiar waxy substance under the microscope, as well as tangles of neurofibrils where neurons used to be. Alzheimer’s boss, Kraeplin was a believer in the idea that psychiatric disease was caused by changes in the physical structure of the brain. When Kraeplin wrote the 2010 version of Psychiatrie, he included Alzheimer’s single patient case study and inflated it into a disease. However, we don’t know if the plaques and tangles caused the dementia, or if the dementia caused the plaques and tangles, or if a third factor caused them both. Also, we don’t know if the plaques and tangles ever occur in people without dementia.

• AD falls under National Institute on Aging (NIA) instead of National Institute on Neurological Disorders and Stroke (NINDS). NIA was a young institute at the time, and the founders realized that aging isn’t something we can avoid or cure, so they advocated for AD to be under NIA to attract more funding. Their argument was that aging is a huge risk factor for sporadic AD, and may be an unavoidable part of aging. To do this, they greatly expanded the meaning of AD to encompass as much of age-related dementia as possible. They backed the amyloid idea of AD and as more and more dementia brains were examined microscopically, plaques and tangles in a characteristic pattern showed up. Now, 2/3 of the annual NIA budget is spent on AD research. Between 2012 and 2020, a big lobbying effort led by Alzheimer’s Association led to more than five-fold increase in AD funding, most of it went to NIA.

But couldn’t they have been acting in good faith on the science at the time, rather than “politically” to call these cases AD? Also the author later says that 1/3 to ½ of us will get AD eventually, so does that mean he agrees with the inflation after all?

• NIA working groups in collaboration with the Alzheimer’s Association come up with a list to determine probable and definite AD. In Definite, the person needs to have dementia, as well as a certain about of plaques per microscopic field. Younger people need fewer plaques to meet the criteria. This was despite the fact that plaques were not present in 15% of people who met clinical criteria for AD. Tangles were listed as helpful but not necessary. They also defined “preclinical” AD as people with normal cognition who had a certain amount of plaques or amyloid in CSF. So now the 30% of elderly people who have plaques but normal cognition are not healthy, but instead have preclinical AD. This means that estimates of AD prevalence went up three-fold.

Biological aging: Aging seems to be not one regulated process, but more the biological consequences of selecting for a lot of genes that help us when we’re young but not when we’re old (antagonistic pleiotropy) or genes that harm us when we’re old but are neutral when we’re young, so they’re not selected against. For example, senescent cells are probably a cancer defense mechanism, but when we’re old, they release inflammatory factors and are bad for us. Also, cells in the body can be highly variable in their rate of aging.

AD does seem to be it’s own process, and not just accelerated biological aging, but there’s some overlap, like increased oxidation, failure of DNA repair mechanisms, inflammation, increase in insulin resistance (which in the brain can trigger neurons to become senescent), etc.

What should we do next?

• Herrup argues that we don’t have a good mechanistic idea of AD at all, and need to return to just a symptom based definition and move forward from there.

• Separate early and late onset AD into type I and type II, because while they share similarities, there are some important differences so studying one doesn’t necessarily translate to the other.

• Presence of plaques and tangles should only be used to calculate risk, not define disease

• View the loss of normal brain function as a distortion of the network of interactions among cell groups, and not a blockage of a single linear biochemical pathway. For example, if we’re testing the idea that the loss of myelin is a major contributor to AD, instead of trying to figure out which cell type is responsible, oligodendrocyte or neuron, focus on the interaction between them. It wasn’t clear to me how this is different than how people are already thinking of neurodegeneration.

• Take all the AD-related funding out to NIA and put it in NINDS. But also, replace all that NIA funding with basic research on aging funding

• NINDS should set aside one third of their review cycles to exclude amyloid or tau based applications. Then research diversity will grow, even within traditional amyloid labs

• We need to stop corporate welfare to pharma. Industry needs to invest more serious money in basic research, and the entire cost of phase III trials needs to be funded by industry with no government or academic money. This way, pharma will be more conservative with phase III trials since they’re fronting the whole cost. But right now only 10-15% of the cost of phase III trials is non-industry, and usually for a non-commercially viable intervention like repurposed generics, vaccines in low income areas, and lifestyle interventions. So would that actually help?

• Universities need to charge companies indirect costs for a scientist consulting for pharma

• Pharma needs to consult not just with scientists in the field of AD, but with smart scientists in peripherally related fields so they get non-group think ideas

• Conferences on AD should mix symposiums to include clinicians, basic researchers, and people in peripheral fields that have ideas relevant to AD, again, to reduce group-think

• Flagship journals need to hold AD research to a higher standard. Too much amyloid-cascade stuff is being published despite weak evidence for the theory

• Science journalists should try to find alternative viewpoints to the amyloid cascade idea

Where the field is now (not from the book):

Most researchers have moved toward a "multi-factorial" model of AD, where:

• Amyloid-beta may trigger or accelerate disease in certain contexts (especially genetically predisposed individuals).

• But other processes like tau pathology, neuroinflammation, vascular dysfunction, mitochondrial failure, and aging are critical co-contributors.

There's also growing interest in:

• Resilience factors (why some tolerate Aβ without symptoms). For example, a rare homozygous mutation in the ApoE3 gene, the Christchurch mutation, led to a woman having the PSEN1 mutation and lots of plaque buildup and high tau pathology, but was cognitively normal into her 70s. Usually, the PSEN1 mutation leads to dementia in your 40s and death by your 60s.

• Microglial response (like TREM2 activation therapies that bind to microglial receptors and make them act in more helpful ways)

• Timing (amyloid-targeted drugs might only work very early, before symptoms)