Dennis Selkoe on the Amyloid Hypothesis of Alzheimer’s Disease

Special Topic of Alzheimer’s Disease Interview, March 2011

According to our Special Topics analysis on Alzheimer’s Disease, the work of Dr. Dennis Selkoe ranks at #1 by total cites and #5 by cites/paper, based on 136 papers cited 19,587 times over the analysis period. In our 2003 Special Topic on Alzheimer’s, Dr. Selkoe ranked at #5 by total cites.

Dennis Selkoe

His record in Essential Science IndicatorsSM from Thomson Reuters includes 136 papers cited 19,806 times between January 1, 2000 and October 31, 2010. These papers can be found in the fields of Neuroscience & Behavior, Biology & Biochemistry, and Clinical Medicine.

Selkoe is the Vincent and Stella Coates Professor of Neurologic Diseases at Harvard Medical School and the Co-Director of the Center for Neurologic Diseases in the Department of Neurology at Brigham and Women’s Hospital in Boston, MA.


SW: You’ve been studying Alzheimer’s for over 30 years, so let’s start with the big question. What’s your working hypothesis of disease causation circa 2011?

My opinion is that there is an imbalance between the production and the removal of a small hydrophobic protein called amyloid beta that triggers the process we call Alzheimer’s. I believe that imbalance arises from a lot of different, more fundamental causes. What I’m saying is that amyloid beta is both necessary and, at least in some cases, sufficient to cause Alzheimer’s disease, but there are many other factors.

If we had to choose one, and I think the clearest, path to treatment, it would be targeting amyloid beta rather than any of these other factors, including tau, which looks like it comes downstream in the Alzheimer’s cascade. So to summarize, my opinion is that it’s an imbalance of amyloid beta protein in the brain that triggers or precipitates Alzheimer’s.

SW: So what causes the imbalance in amyloid beta?

When you use the word “cause,” it gets difficult. We have lots of genetic evidence that this amyloid beta protein imbalance triggers Alzheimer’s—that’s one reason that my work and others is cited a lot—but if you then ask what imbalances the amyloid beta, well, that’s more complicated. Certain gene mutations imbalance its brain levels dramatically, and they cause very aggressive familial forms of the disease. Those are rare, though, and people have often criticized the amyloid hypothesis, saying that those are really rare forms, and we don’t have compelling evidence that conventional Alzheimer’s follows along the track of familial Alzheimer’s.

“I’d like to focus on exactly how the amyloid beta protein first binds to cells and injures them. We don’t know how that happens.”

I don’t agree with that. When I see patients in the clinic, and I still do, those with familial Alzheimer’s look the same as typical Alzheimer’s patients who don’t have a clear family history of the disease and for whom there is no known genetic cause. When you look at the brain of a familial Alzheimer’s patient and a so-called “sporadic” patient, they look indistinguishable. No expert can tell them apart. That’s good evidence that the common form copies the rare familial form. But there’s a lot more evidence than that.

SW: Can you give us a brief tour of the other evidence arguing for amyloid beta protein as the fundamental trigger of the disease?

One of the most striking pieces of evidence is that in the spinal fluid of essentially all people who develop Alzheimer’s, amyloid beta levels go down even before they develop symptoms. This has been tested and observed around the world—in the US, Europe, Japan, China.

You see this decline in the soluble form of Abeta42 (the more self-aggregating 42-amino acid form of amyloid beta) long before the patients have symptoms. It could be three, four, even five or more years before symptoms become apparent. This lowering is due to Abeta42 protein becoming tied up on brain membranes. It sticks to the membranes of nerve and glial cells in the brain. It doesn’t float in the interstitial fluid of the brain, and consequently the spinal fluid level reflects that sequestration. We have recent preclinical (mouse model) evidence for such a mechanism explains the tell-tale drop in Abeta42 in the cerebrospinal fluid (CSF).

If you ask, “Can we see the buildup of amyloid in a patient’s brain?”, well, the breakthrough work from Bill Klunk and Chet Mathis at the University of Pittsburgh in 2002 showed that you could put a radiolabeled dye into the body that will cross the blood-brain barrier and bind to amyloid plaques, emitting a small radioactive signal seen on a PET scan of the head. So we can now see that as CSF Abeta42 falls, it does so while less soluble amyloid (as seen by PET scanning) is beginning to build up in the brain. If you then ask, “Which of these two markers is even more sensitive to the preclinical stage of Alzheimer’s disease than the other, before symptoms begin,” it would be the spinal fluid level.

Another example of the apparent primacy of amyloid beta, one of the most powerful examples, is Down syndrome. That is caused unequivocally by an extra copy of chromosome number 21, and essentially 100% of people with three copies instead of two go on to get Alzheimer’s disease. They get the neuropathology of Alzheimer’s, and they actually get confusion, forgetfulness, and disorientation during young to middle adulthood, beyond whatever cognitive dysfunction was already there during their childhood and teenage years.

“We have lots of genetic evidence that this amyloid beta protein amyloid beta imbalance triggers Alzheimer’s—that’s one reason that my work and others is cited a lot—but if you then ask what imbalances the amyloid beta, well, that’s more complicated.”

You can say, well, there’s a lot going on in chromosome 21. A whole extra chromosome, some 2,000 genes or so. But sure enough, some astute clinical researchers followed a woman who had a translocation form of Down syndrome, in which chromosome 21 is broken and a piece is translocated onto another chromosome, and that led to her having Down syndrome. She had the extra genetic material from that broken piece of chromosome 21, but the doctors noticed that when they mapped that chromosome, the breakpoint in her chromosome 21 was “south” of (telomeric to) the amyloid precursor protein (APP) gene. The extra piece of chromosome she had in all of her cells did not include the APP gene.

That’s very unusual, and they followed that woman for 20 years or more, and she never seemed to be much different cognitively in her old age than she was when she was younger. Eventually she died in her mid-seventies, I believe, and she had nary a plaque of amyloid in her brain, which is in striking contrast to almost every other patient with Down syndrome. I thought that was a beautiful piece of clinical research that pinned down the notion that if you don’t have an extra copy of the APP gene when you have Down syndrome, then you won’t get the classic neuropathology of Alzheimer’s in the brain. That was very instructive.

I could go on and on, but I feel very comfortable with the notion that we should try to treat Alzheimer’s by targeting and safely lowering the amyloid beta protein. And if you ask the question why our research is so highly cited, it’s because we’ve been doing a lot of work over a long period of time, since the late 1970s, showing how amyloid beta is made and how “it does its dirty work,” in a simplistic expression.

SW: In a 2003 Q&A for Special Topics, you said that the presenilins are absolutely central to the pathology of Alzheimer’s. Is that still considered to be the case or has the science changed significantly in the intervening years?

It’s very much the same or even amplified. Presenilins are absolutely essential in this disease. This is the enzyme that actually makes the amyloid beta protein. Everybody has that enzyme normally. Everybody is making a little amyloid beta throughout their lives, as my lab discovered in 1992. Seven years later, with my colleague Michael Wolfe, we discovered that presenilin was the enzyme that made amyloid beta. It was just referred to as gamma secretase heretofore.

There’s another enzyme, beta secretase, that cuts APP first. Then, presenilin (aka gamma secretase) cuts the remaining piece of APP to release the amyloid beta protein. Imagine you have a piece of string and you want to release a smaller piece from the middle—you need to cut it twice. If the original string (i.e., APP) is like the alphabet, running from a to z, then you cut first at m and later at q, and you get a little piece that goes from m to q. Well, one cut in APP is made by gamma secretase (at q) and the other is made first by beta secretase (at m). This gamma secretase enzyme is now called the presenilin/gamma secretase complex, and it’s just as important as we thought it was eight years ago.

You can’t get Alzheimer’s without presenilin cutting APP. The presenilin gene and its protein turn out to be necessary to get Alzheimer’s. And it turns out that though it was discovered through the study of AD, presenilin is relevant to many, many other normal biological events and even to other diseases. This is because presenilin/gamma secretase has more than a hundred different substrates, which are proteins that it cuts. Only one of them is APP.

SW: Why do you think there’s been so little progress, if any, in developing a drug that can slow the progression of Alzheimer’s?

That’s a crucial question, and I think about it all the time. I’m actually working toward the improvement of clinical trials and how to analyze them in Alzheimer’s that will help address this. My opinion is that the drugs that have been tested so far were highly flawed.

Let me give a very concrete example: Eli Lilly had one of the biggest disappointments in Alzheimer’s treatment when it stopped a phase III trial of a new drug last September. Their drug was a gamma secretase inhibitor, which is just what I’ve been saying I’d like to see. But it was not a good drug. It inhibited presenilin/gamma secretase from cutting many, many different proteins—other proteins—and one in particular is called Notch, which is a very famous molecule in biology. It’s very important in fruit flies, worms, humans. It controls aspects of cell fate; how one cell becomes one thing and another cell becomes something else entirely. And you don’t want to mess with that, even in adult humans, where it’s important in the gastrointestinal tract and bone marrow. For example, to make the proper cell in the gut that makes acid, we need Notch.

Now, Lilly’s drug had what we call a therapeutic index of 3. That means the dose of the drug that does the good thing—blocking the cutting of the APP by presenilin to lower amyloid beta—is one-third that of the dose needed for doing a bad thing—inhibiting the cutting of Notch. At first glance, that might not seem so bad; you need three times the activedose before you block Notch. But it still isn’t a good therapeutic index, and frankly a number of scientists in the field said in advance that it doesn’t look like a good drug to test. God forbid it doesn’t work; it will set back the field. People will become disillusioned. Or God forbid the patients get sick.

“My hypothesis, and I haven’t written this widely, is that amyloid beta protein does not bind to a particular protein receptor on cells.”

Well, Lilly announced in September that the patients didn’t get better and they stopped the trial short and said some people in the trial had clear signs of Notch toxicity, for example, developing a type of skin cancer. This is not a minor manner. And they had other problems—gastrointestinal problems, etc. Lilly also said, by the way, when they cognitively tested patients during the trial, they actually seemed to do worse on the drug rather than better.

SW: So what did that say about the amyloid hypothesis itself? You certainly don’t seem to have lost faith in it.

People uncomfortable with the amyloid hypothesis said they knew this would happen: amyloid is the wrong target. You don’t want to lower it. But the fact of the matter is that it’s far more likely that the multiple side effects the patients experienced with Lilly’s drug led to somewhat poorer cognition while they were on the drug.

When my patients with Alzheimer’s disease get urinary tract infections, which they do, their dementia seems to be much worse. They get more confused because they can’t tell you that it hurts a little bit when they have to go the bathroom, and they have to go a lot. Their children call me and say “suddenly, my mom is just different.” I say take them down to the doctor and get a urine culture, and when they take care of the urinary tract infection the dementia improves again.

And my point is that in Lilly’s trial, they probably caused enough side effects by inhibiting the cutting of Notch protein and, indeed, other proteins also, so there was a price to pay. Patients got more confused. If I’m right about that, then when they follow these patients longer and the Notch side effects clear up without the drug, they should go back almost to their baseline cognition, or perhaps just a little worse, given the time that has passed.

SW: Are you discouraged by the progress that’s been made, particularly after the failure of this Lilly drug?

This Lilly drug is now the most common failure cited as evidence that the Alzheimer’s field is in a state of malaise, that it’s really in trouble and we may not know what we’re doing. I don’t think that’s right at all. We’re moving right along with experimental treatments. There are about 60 or so trials ongoing for Alzheimer’s drugs worldwide, some in phase I, some in phase II, some in phase III.

The problem is that the first three drugs that were tried in large trials were all poorly thought out. People, understandably, were anxious to get their companies into drug trials for Alzheimer’s disease, and they seemed to have chosen the wrong drug. The other two examples that also failed in phase III, like the Lilly drug, had poor attributes as drugs. They were unlikely to be safe and effective.

To summarize, I believe the field is moving along in a rational manner. Therapeutics are based on what we know about the biology of a disease, and companies and academic researchers are developing agents, doing so on a firm footing—although it’s never firm enough—and doing the necessary trials. And I think the current trials are going well enough. Some of these drugs have the chance to be successful, and it means they might really help the patient and we’ll know more in the next year or two. But they will only be successful if they are tried in patients with mild Alzheimer’s disease, not moderate or advanced, when it appears to be too late to really have an impact.

SW: What would you like to accomplish in the next five years?

I’d like to focus on exactly how the amyloid beta protein first binds to cells and injures them. We don’t know how that happens. We know, with all this genetic and biomarker and pathology evidence briefly alluded to above, that amyloid beta protein builds up as you get Alzheimer’s, but we don’t know exactly how it short-circuits the nerve cells. We have some very exciting experiments underway to figure that out.

“…my opinion is that it’s an imbalance of amyloid beta protein in the brain that triggers or precipitates Alzheimer’s.”

My hypothesis, and I haven’t written this widely, is that amyloid beta protein does not bind to a particular protein receptor on cells. I don’t think there is a specific receptor for amyloid beta, and certainly not for the amyloid beta doublets and triplets (“oligomers”) that appear to be particularly noxious.

Those don’t have a natural receptor, and I think instead they bind to a fatty substance on the nerve cell membrane, a lipid. The amyloid beta protein doublet (dimer) is very hydrophobic. It hates water. It has some of its most hydrophobic amino acids sticking out, ready to accept another amyloid beta protein, so that a dimer goes to a trimer and on to a tetramer, etc., and it suggests that amyloid beta protein assemblies will do anything they can to get out of an aqueous environment. They can’t tolerate water. So they stick to the very sticky lipid environment, which is the surface of these nerve cells and many other cells in the brain.

Another question that I really would like to understand is how antibodies can be used to prevent or treat Alzheimer’s. You’re probably aware that advanced trials are underway—they’ve not yet failed, and I hope, of course, that they won’t—of vaccines and antibody treatments for Alzheimer’s. I would like to know more about how the latter might benefit the patient.

My colleagues and I wrote a paper in Nature Medicine in 2008, which has been cited a lot, describing, among other things, how certain antibodies can neutralize the bad effects of amyloid beta dimers on nerve cells (Shankar GM, et al., “Amyloid beta protein dimers isolated directly from Alzheimer’s brains impair synaptic plasticity and memory,” 14[8]: 837-42, August 2008). The dimers (and other, larger oligomers) can cause interference with synaptic transmission.

I would like to know more about how the antibodies prevent that, because in my biased opinion, the best shot we have now is immunological treatment for amyloid beta, an antibody against amyloid beta, which is currently in a phase III trial in some 25 countries. I would like to know more about how that antibody neutralizes the effects of amyloid beta protein and indeed whether it actually does so in humans.

I’d also like to find out more about proteins that misfold in other brain diseases. One my lab is studying now is the protein alpha synuclein, which I think is going to turn out to be for Parkinson’s disease what amyloid beta protein is for Alzheimer’s. We’re working on how that protein misfolds and aggregates and causes trouble for neurons, although it’s going to be a somewhat different mechanism than how the amyloid beta protein causes trouble in Alzheimer’s.

SW: Of all the research in Alzheimer’s done in the past five years, which experiments did you read about and think, “I wish I had done that?”

I have to give more than one example. I can’t elevate one over several others. So, one would have been the discovery that a duplication of the APP gene in otherwise normal people that can precipitate a very nasty form of Alzheimer’s. That discovery was made in France about six years ago.

I thought that was very clever because we knew that mutations in the APP gene could cause Alzheimer’s and we knew that trisomy 21 could also cause an Alzheimer’s phenotype, but these folks in France did some clever genetic sleuthing to find rare families that had an extra copy of just the APP gene and maybe two or three other genes right around it on chromosome 21. These people were physically and mentally normal throughout their lives, until they got Alzheimer’s. That was a great smoking gun for the amyloid beta hypothesis. I would have loved to be associated with that.

Another discovery that I admire greatly was done by Lennart Mucke at UCSF. He got some mice that had the tau gene deleted genetically, and he crossed those “tau-minus” mice with mice that had the human APP gene inserted into their genomes. He showed that while amyloid beta protein still built up in the brains of these crossed mice, they had far less behavioral problems than the regular APP transgenic mice with tau. In other words, tau played a big role.

Tau is a subunit of the Alzheimer’s neurofibrillary tangles. And Mucke and his colleagues were able to show that if mice don’t have tau, they still get the amyloid pathology of Alzheimer’s but they don’t get all that much behavioral trouble. They have behavioral symptoms but much less. I thought that was really cool, and I would have enjoyed coming up with that myself.

“…my lab is studying now is the protein alpha synuclein, which I think is going to turn out to be for Parkinson’s disease what amyloid beta protein is for Alzheimer’s.”

We did follow up on it in our laboratory with a paper that’s almost in press (I hope) at PNAS. I’m very excited about it. We said, “Let’s reduce Mucke’s discovery to an even simpler form.” We took some primary cultured neurons from rats and put them in a dish and then put on top of them some soluble amyloid beta dimers that we isolated from the brains of Alzheimer’s patients after they died. The idea was to see if those amyloid beta dimers are themselves necessary and sufficient to induce alterations of tau. And we showed it quite nicely. When we put on these amyloid beta dimers, even in exquisitely small amounts, they were very potent.

First we induced an increased phosphorylation of the tau protein in these healthy neurons. Then we saw the microtubule cytoskeleton begin to collapse, and then the nerve endings degenerate into what we call neuritic dystrophy. So, in a test tube, a culture dish, we can show how AD brain amyloid beta protein dimers directly induce these tau alterations, the abnormal phosphorylation of the tau protein, just about the same as happens in the Alzheimer’s brain.

Then, when we knocked down tau in the neurons using RNA interference, we rescued the neurons and they didn’t go on to degenerate. That was really cool. And we also showed that if you treat the neurons with the very same antibody that’s in phase III clinical trials, it prevents the bad effects of the amyloid beta protein dimers on tau and on the neuritic architecture.

We also learned in this work that the neurons have to become more mature, somewhat older, to see these effects. If you take neurons that are just seven days in the culture dish, they don’t have any of the bad effects from the amyloid beta dimers. Only after 14 or 18 days in culture or, best of all, 21+ days, will the neurons have the right protein expression program underway to really suffer from the amyloid beta dimers. Before that, they’re resistant. Biologically, that’s very interesting. I like to think that this experiment shows the bridge between the two classical lesions that Alzheimer’s first described in 1906—plaques and tangles.

SW: If we lived an ideal scientific environment and you had unlimited resources to do a single experiment, something you couldn’t afford now, what would you do?

I would design a clinical trial of 18 months’ duration, in which we would take only people in the very mild stage of Alzheimer’s, people but who do have the disease, who do have symptoms of memory trouble, but are in the early stage. This wouldn’t be a prevention trial. We would use extensive scanning of brains with Pittsburgh compound B or a similar dye to find these subjects—that’s expensive, perhaps about $3,000+ per patient, and we’d want at least 500 patients. Then, we would have to test their spinal fluid for amyloid beta protein 42 levels, which should be decreased.

But the main point is to give the treatment in the early stages of this disease and to enrich the trial population with subjects that we know all have amyloid deposition in their brains. We don’t want to treat people with another kind of dementia, because amyloid drugs won’t help those people. So we get 500 or perhaps even 800 to 1,000 patients, we give a third of them placebo and the other two-thirds one or another dose of the treatment. If you ask which treatment, I’d say probably an antibody to the amyloid beta protein, because we know the most right now about how antibodies work. That would be the experiment. It would be complicated and very expensive, but it’s the experiment I’d most like to do.

SW: How expensive is very expensive?

If we’re doing a phase III trial, it would probably cost in the neighborhood of $100 to $200 million or more to do it right. We’d have to scan the people, follow them carefully, pay for all the costs incurred to undergo the experimental treatment, the doctors’ and nurses’ visits, etc. We’d have to do the spinal fluid exams. Probably north of $200 million. That’s not a laboratory experiment, but if you ask me what I would most like to do as a scientist, as someone who’s worked on this for so long, that would be it. I’d like to design a clinical trial that would be even better than the ones underway right now. I’d like to see this disease defeated.

Dennis J. Selkoe, M.D.
Harvard Medical School
Brigham and Women’s Hospital
Boston, MA, USA