Posted 8 January, 1999
Interviewed by Aileen Anderson
We invite you to submit comments and questions on Dennis Selkoe's remarks.
Q. What is the primary hypothesis that drives your lab group?
A. The primary hypothesis is that a chronic imbalance in the production versus the clearance of A-beta leading to a gradual rise in its steady state levels in brain tissue is the cause of Alzheimer's disease.
Q. In that context, what molecular and cellular changes account for the initial symptoms of AD, and what might be the earliest pathological changes? In the framework of your hypothesis, does A-beta deposition come first?
A. Yes. A-beta accumulation comes first. In view of current knowledge, let's change the word deposition to accumulation, which precedes deposition. Our hypothesis posits that A-beta accumulation or rise in steady state levels is the first event. I refer to A-beta the peptide, which is not amyloid, as we know. A-beta and amyloid have to be kept distinct. I have no idea whether amyloid plaques will turn out to be critical for pathogenesis; I think they probably play a role, but I think accumulation of the A-beta peptide is critical. Thus, I think that the fundamental basis of AD involves a rise in steady state A-beta levels, first as a soluble peptide in the brain interstitial fluid (and perhaps also intracellularly), then as an insoluble oligomeric species, and then going on to a polymeric species which we ultimately call amyloid fibrils. So the earliest pathological changes, if the word pathology is defined classically as what is morphologically apparent in the tissue by microscopy, would be the A-beta diffuse plaque.
Q. What do you think about intracellular A-beta?
A. Intracellular A-beta clearly exists. We've known since our lab and that of Steve Younkin discovered A-beta secretion in early 1992 that A-beta was made by normal intracellular processing of APP, and so A-beta the monomer clearly exists inside cells. We had a poster by Dominic Walsh here at the Society of Neuroscience meeting (1998) that showed that we can detect an apparent A-beta dimer inside cells that overexpress APP, and Virginia Lee's lab has nicely shown that there is A-beta accumulation inside neurons in a form that requires solubilization in formic acid. Therefore, A-beta accumulation and, apparently, oligomerization begins intracellularly. As to whether intracellular A-beta has an important pathogenic role in AD, it does in the sense that it is the source of all the A-beta that accumulates in brain. It all starts intracellularly. Whether it's critical for causing neuronal dysfunction and degeneration, however, is a different question. If you look at the morphological evidence in the disease itself, you would say no. There is no good morphological evidence to date of any appreciable amount of immunogold-detectable A-beta inside neurons or other cell types in AD brain tissue. Even at the end stages of the disease, when there are huge amounts of A-beta readily detected extracellularly, you can't detect A-beta inside cells. So I would say that the chance that it will turn out in the fullness of time that intracellular A-beta has a direct neurotoxic role is low, although I think it is possible that A-beta starts to accumulate and oligomerize intracellularly. If someone ultimately shows that A-beta dimers are the key pathogenic moiety in AD and without dimers one can't make plaques, and on this basis argues that intracellular A-beta is critical, I would respond that I am not arguing against this notion here. Rather, I'm saying that I don't think A-beta accumulates to pathological levels intracellularly enough to destroy neurons from within.
Q. How much of a role does genetics play in getting AD, and will genetics eventually explain all cases of AD?
A. I think genetics plays a very strong role in AD, and that it will turn out to explain a large fraction of AD. I don't think it will turn out to explain 100%, although I can't exclude that possibility. I would guess that in 20 years or so, we will be aware of a large range of genes that are linked to AD, some of which will have dominantly transmitted mutations in them, others of which will simply be polymorphic and confer an increased risk of the disease, like Apo E4 and perhaps alpha-2 macroglobulin. However, as in the last question, the answer depends on exactly how one defines one's terms. If I tell you that at age 85, a woman develops AD with absolutely no elicitable family history, none whatsoever, and she has no ApoE4 allele, no alpha-2-macroglobulin alteration, no other known genetic risk factors, can I then conclude that that hers is truly sporadic (non-genetic) disease? Absolutely not. So then, what are we saying when we ask whether genetics will explain all cases of AD? Is there a pathogenic agent in the environment that can cause AD in many different individuals regardless of their genotype? If so, that would be a true example of a non-genetic form. So let's say there was a very abundant respiratory pathogen in the hotel ventilation system, then virtually all exposed residents might come down with pneumonitis. I doubt we'll find such a clearly environmental cause for AD that works essentially without regard to genotype. Several respiratory pathogens can infect and even kill people worldwide regardless of genotype. If that's what we're looking for in AD, I don't think we'll find such a purely "non-genetic" cause for AD. So, in this sense, I believe that the development of AD will always depend to some extent on internal biochemistry, dictated by the 100,000 or so gene products expressed in the body.
Q. What role does aging itself have in AD pathogenesis?
A. The way I would think about that would be to suggest that aging has at least 2 broad roles. The first is that aging is time, time on the planet. AD is a biochemical abnormality in the accumulation and aggregation of a small hydrophobic peptide, and we know from studis in vitro, in transgenic mice and even in humans, that it takes time for A-beta to accumulate sufficiently outside neurons to cause neurotoxicity. So in this context, "aging" is simply "time"; that is not the usual sense that one uses the word. When a 35 year old woman presents clinically with mutant presenillin-caused AD, she is not old by most conventional criteria, and yet she gets a devastating brain disease and may die of the disease by 43. The reason, of course, is that she has a mutation in (I believe) gamma secretase (or perhaps in a critical cofactor for gamma secretase), so that it only takes her 35 years to build up enough cerebral A-beta to initiate symptoms, instead 70 or 80 years. So I think that the role for aging is time, first and foremost. The phenotype (both clinically and neuropathologically) of that 35 year old woman is remarkably similar to that of a 75 year old "sporadic" AD case. So you do not require aging, per se, to get the full complement of brain abnormalities that we call AD. Having said that, I think that aging plays a role in AD pathogenesis in most cases, because most people don't have this type of aggressive mutation, and therefore they take a lot more time to accumulate their peptide and allow it to cause trouble. That's where the second meaning of aging comes into play. When this disorder occurs in someone who is 75 rather than 35, there are many systems that are not as functional in the latter individual, and so there is probably an acceleration of the A-beta-driven impairment both pre-clinically and after symptom onset, because older neurons (and perhaps glia) can't handle it -- there is a limit in their capacity for compensation and repair.
Q. So your idea is essentially that any genetic vulnerability predisposing to AD that that 75 year old might have will be accentuated by additional vulnerabilities that develop in the aged brain?
A. Exactly right. The onset and course of the disease could depend on how you "handle" your plasma membrane in your neurons, let's say, or how robust your synapses are. In that sense, again, I would say that I don't think aging per se or advanced age is a prerequisite for the disease, because you get full blown, typical AD in a young individual, but I think the aging process often hastens the onset of clinically detectable disease.
Q.Can you outline the progression that you expect AD takes, and why does the first
stage of the disease follow a particular anatomical pattern?
A. I think that the progression of AD begins with some change in the "economy" of A-beta, that is, some imbalance between the production and clearance of this peptide. Then, when A-beta is present at sufficiently high levels, presumably in the interstitial fluid of the brain extracellularly but maybe to some degree intracellulary, it then triggers its own oligomerization, probably with the help of some kind of chaperoning proteins. Once oligomers are formed, they may exert toxicity, or they may form a nidus for even more oligomerization or polymerization towards the occurrence of a diffuse plaque, which is largely non-fibrillar. The next event, I suspect, is the accrual to the A-beta of other soluble molecules (proteins) in the interstitial fluid, and perhaps almost simultaneously, the activation of local microglial cells. Thinking in terms of what cell type could most readily respond to the presence of "foreign" material, which I believe A-beta oligomers and polymers represent as far as the host is concerned, microglia are probably more able to respond rapidly than neurons or their synapses are. So I would favor the sequence: A-beta accumulation, oligomerization/polymerization, microglial activation, cytokine release, astrocytosis, and neuronal/neuritc alteration secondary to the microglial and astrocytic activation, including a large array of released proteins from the latter cells. These things could all occur in one virtually simultaneous response. There could be direct neuronal/neuritic injury from A-beta, I can't exclude that, and presumably it's all so very complex and packed into a very small interstitial space that alot of things are going on at once. And that's why I think that neuronal/neuritic injury, per se, is not a good therapeutic target. Then downstream of that, I think there is change in neuronal calcium (and other ion) homeostasis. There's also secondary oxidative injury to internal molecules in the neuron. I think the neuron is perturbed from the surface, and it could be perturbed by cytokines, it could be perturbed by free radical generation by microglial cells, a large variety of things. And then, of course, the process marked by tangle formation is important. Other downstream steps include progressive synaptic dystrophy, frank neuronal loss, neurotransmitter deficits, and finally clinical dementia.
Q. So how important do you think activation of the immune system is? E.g the complement sufficient/insufficient mice story for the APP transgenics?
A. If we're talking in terms of the antigen presenting cells in the brain, I believe the microglia and their local cellular response elements are probably going to turn out to be very important. In my opinion at this time, the microglial response may well come before neuronal/neuritic alterations.
Q. So do you think that neuronal/neuritic alterations require a microglial/inflammatory response?
A. Probably. Well you can't say for sure, but since these events are very difficult to separate in vivo, it becomes virtually an absurd question. Since we'll never have a human host that doesn't have a microglial and astrocytic response, I think it doesn't make any difference. In culture, you can injure neurons directly. In vivo, I suspect you can't have neuronal injury without also having microglial and astrocytic alterations.
I believe the second aspect of your original question was the anatomical specificity of the disease. That's a very difficult one. I think that the anatomical/cytological phenotype that develops in AD is driven almost exclusively by the local response to A-beta. In other words, where A-beta is able to accumulate into diffuse plaques and ultimately into so-called mature fibril-bearing plaques is a highly regulated process in the brain. We know in striatum and cerebellum, very few hosts develop fully mature plaques, even though they may have a lot of diffuse A-beta deposits there in some cases (particularly in the striatum). And yet in that same host, the hippocampus, entorhinal cortex, amygdala and association neocortex will develop many more mature plaques. Therefore, I think that the regional specificity of the disease is conferred by so far unknown locally produced factors. We might already know the names of some of them, but we don't yet know that they actually have this function. I am speaking about factors, presumably proteins, that enable excess levels of A-beta monomer to gradually transmutate from monomeric to oligomeric to a fibril-rich highly polymeric form. .I think that it's not a coincidence that where the mature fibrillar plaques are, that's where one sees microglial activation, astrocytosis, neuritic surround, synaptic degeneration. So I think that regional specificity is conferred by still ill-defined factors which promote the slow maturation of diffuse to mature plaques. At this initial level of anatomical specificity, it's not a regionally-specific vulnerability of neurons that we are speaking about. The latter represents another way for anatomical specificity to be conferred, but we do not need to invoke that initially. Before we even worry about whether one neuron does or does not respond to aggregated A-beta, we need to deal with a dramatic difference in the regional maturation of A-beta-containing lesions between areas that are clinically "spared", like the striatum, and clinically affected, like hippocampus.
Q. Do you have any speculation in the context of the reports that the mice, certainly the APP/PS1 transgenic crosses, progress somewhat differently in terms of where you first see A-beta deposition in the human AD brain.
A. I don't agree with that statement. I think that several of the reported mouse models, while they may actually develop their very first deposits in the cingulum, usually develop the most abundant initial deposits in the hippocampus. In fact, Cindy Lemere in our Center has just studied this quite intensely using the PDAPP mouse, in collaboration with Dora Games and her colleagues. In those mice (solely overexpressing APP), they have definite early A-beta deposits, diffuse and later more mature looking, in the hippocampus and in the molecular layer of the dentate gyrus. I used to believe that hippocampal lesion formation per se wasn't the "Holy Grail" of AD pathogenesis, but the human APP transgenic mice have convinced me that AD-type neuropathology begins in the hippocampal and entorhinal structures. I think most in the field believe that there is a very early and profound hippocampal disturbance, therefore explaining the patient's problems with retention of new information and decline in the most recent memories. I think the mice generally reproduce this quite beautifully; almost always in the mice I have studied, the hippocampus is affected before the frontal cortex and just about at the same time as the cingulum is.
Q. My understanding was that the APP/PS1 cross exhibits pathology in the frontal cortex first.
A. Some people have said that it affects hippocampus early and some people have said that it is in cingulum or frontal cortex, but in the mutant APP transgene by itself, hippocampal pathology usually precedes other areas. It can also be, as Karen Hsiao first emphasized, that neuropathology is strain specific as well.
Q. How about reports that in the canine frontal cortex deposition precedes other areas?
A. This finding could certainly support the notion of species differences. But in humans, I suspect that hippocampus and amygdala and other limbic structures are particularly vulnerable to the "AD process", whatever one believes that process is, and my hypothesis would be that it's related to the accumulation and maturation of A-beta deposits to a more toxic form.
Q. At what stage in the event pathway for AD do you see intervention being most successful?
A. Intervention will be most successful at the earliest possible stage, and that, to my knowledge, is when A-beta is released from the neuron or when A-beta is generated, which is even earlier. And so, I would interfere with the generation of A-beta. I think that's a scientifically rational hypothesis. It beautifully follows the example of the role of cholesterol in the pathogenesis of atherosclerosis, in the sense that if one chronically lowers cholesterol in the serum in healthy, asymptomatic individuals, one can retard the development of lesions and decrease the risk of developing clinical cardiovascular disease. Cholesterol modulation has started to be successful for people at risk many years or decades before symptom onset. In the same way, I believe that chronic gamma secretase inhibition would be the most sensible way to get both a preventative and a treatment for AD.
Now, are there disadvantages to it? Of course; you're interfering with a constitutive enzyme activity. But in medicine, we have an excellent experience in chronically inhibiting enzymes in humans and getting away with it: angiotensin converting enzyme in the case of Captopril, 3-hydroxy-3-methylglutaryl coenzyme A (HMG CoA) reductase in the case of Lovastatin, HIV proteases in the case of Saquinavir. There are several drugs on the market that chronically inhibit proteases or other enzymes, and patients tolerate these relatively well, enough to take the drug chronically. I think we now know enough about the biology of APP inside the neuron and other cells to interfere safely with the gamma secretase. This would be an attractive choice, because then you would affect neither alpha- or beta- secretase cleavage. The best of all might be to have a selective inhibitor of that gamma secretase activity which generates A-beta42 and p342 and thus theoretically leave untouched about 90-95% of APP metabolism. Of course, there are several other amyloid-based therapeutic interventions that I have summarized elsewhere, but this (gamma secretase) is the first point of attack, in my view.
Q. So if we had that drug and we brought it to clinical treatment, would our strategy be similar to what we do with cholesterol? Would we assay A-beta levels, look at your genetic history and say, ok, you need to be on this inhibitor for the rest of your life?
A. Absolutely. Ten years from now, it might be 15 years, hopefully not 20, physicians will perform in their office a thorough AD risk assessment profile that will include genotyping for those genes that might be indicated, based on the nature of the family history. This wouldn't be done routinely, but certainly would be if there is a family history of AD (or any AD-like clinical dementia if no confirmatory autopsy data are available in the family). And even for the predisposing genes that represent polymorphisms, like ApoE and perhaps alpha-2-macroglobulin (where its difficult to say in any one family whether AD is clearly inherited), we probably would check for these anyway. We might do this genetic risk assessment at age 55 or 60, for example. And then we'll also measure plasma A-beta 40/42 ratios and in appropriate individuals at risk, CSF A-beta 40/42 and Tau levels. Then, on the basis of these results, the physician will calculate a numerical index that will tell one what the risk of developing AD is, having studied these various parameters in large numbers of people worldwide and taking into account the subject's ethnic group. Gathering these epidemiological data will be time consuming, but it should not be intellectually difficult. We will assemble the most useful genetic and biochemical risk factors, screen several populations for them, and decide to what extent a certain profile of tests predicts the occurrence of clinical AD in a population. If you have a particular numerical risk of developing AD sometime in the next two decades, for example, you may be offered treatment with a gamma secretase inhibitor. If you have a borderline or low risk, you probably would not be offered the drug, unless the inhibitor is entirely without side effects.
Q. If it turns out that you are wrong about what the cause of AD is, what other hypothesis is your favorite?
A. Great question. But I have to ignore an enormous amount of genetic, clinicopathological, biochemical and animal modeling data to come up with an answer. So, if I'm forced to answer the question, I'd have to say what's wrong is that the mutations in presenilin and APP and the ApoE4 polymorphism all result in other neuronal (and/or glial) abnormalities and they affect A-beta generation or levels as a secondary or tertiary event to those other, as yet unknown misfunctions . The dysfunctional proteins, for example, change the internal environment of the neuron (or glial cell) and destroy the neuron from within. In this scenario, the A-beta part of the pathology would be secondary or tertiary, it might contribute somewhat to cell injury or it might be innocuous. That hypothetical possibility is very difficult to imagine, because then you have to require presenilin, APP, and ApoE4 to have a similar functional role in neuronal or glial metabolism somehow, because the resultant clinicopathological phenotype is quite stereotyped among these different genetic causes of AD. My opinion is that while it's fun to try to answer the question, it's almost too late to ask it in a compelling fashion, unless most of our current biological understanding is grossly incorrect.
Q. What is the primary piece of evidence that would convince others that your central hypothesis is correct? Given no technical or funding limitations, design the perfect experiment to prove your idea.
A. Easy. The obvious way to prove the hypothesis is to take a cohort of 400-500 patients with mild or mild-to-moderate clinically typical AD, divide them into 2 thoroughly randomized groups, treat one group with a gamma secretase inhibitor chronically for 24 months, and the other group with a placebo, and otherwise manage the patients identically during the trial. Measure A-beta 40 and 42 in the plasma at 2-3 month intervals during the course of the trial (and perhaps CSF A-beta and tau levels at entry and at 1 and 2 years in a subset of patients). Do regular cognitive testing with a broad array of psychometric instruments. Do SPECT scanning both before entry into the study and at 6 month intervals during the trial, and consider also following volumetric MRI of the hippocampus. Closely follow for 24 months the clinical progression of the disease, as well as SPECT scans and other surrogate markers. In the ideal trial with unlimited funds and study approval, you would perform a lumbar puncture before entry to make certain that the A-beta42 and tau ratio confirms a diagnosis of AD, just taking those who have a positive test and leaving out others, since almost no one would dispute that those who clearly have low A-beta42 and high tau almost certainly have AD. As an arm of such a trial, one could enroll a cohort of patients with mild AD and a presenillin 1or 2 mutation. You would closely measure the psychometric and behavioral progression of the disease in the 2 treatment groups, follow the surrogate markers, assure that plasma A-beta40 and 42 levels are chronically decreased (hopefully by about 40-50%) throughout treatment. The results would help support or refute the hypothesis that chronic elevation of A-beta levels in the brain underlies AD. Only in human trials can the hypothesis ultimately be proven; that's the only definitive data that will convince us. Further preclinical experimental work, including extensive trials in transgenic mice will add icing to the cake, but will never prove the case. You have to have some substantial stabilization of the course or actual improvement in symptoms (or else delay or prevention of the onset of new disease) to confirm the A-beta hypothesis.
Q. What is the most nagging criticism that is raised against your primary hypothesis? What is the main distraction, the thing that would tend to lead people away from your central point?
A. Perhaps the most nagging criticism that one hears over and over again is that A-beta burden and plaque counts correlate poorly with the presence and/or the degree of clinical impairment. But recent studies, for example the nice quantitative analyses of Brian Cummings (Neurobiology of Aging, 1997), refute this strongly-held assumption, as do some of the very early plaque and tangle quantitation studies of Blessed, Tomlinson and others. There exists convincing data that both total A-beta-immunoreactive deposits and neuritic plaques do show statistically significant and sometimes very strong correlation coefficients with degree of dementia. But obviously, such clinical-pathological correlations simplify a very complex phenotype. Since diffuse plaques often occur (sometimes abundantly) in cognitively normal older individuals and these lesions are widely believed to be precursor lesions, we must expect a complicated relationship between A-beta amount in the brain and clinical impairment. Some hosts can apparently tolerate quite high cerebral burdens of monomeric and oligomeric (diffuse) A-beta with minimal, "sub-symptomatic" glial and neuronal/neuritic alterations. This is just as one would expect from considering the complex, multifaceted relationship between vascular cholesterol deposition and clinical vascular syndromes (angina, MI, CVA). Slow, chronically evolving pathologies in aged humans generally do not have a simple, linear relationship to symptom burden. This fact simply does not argue that the early lesions (fatty streaks of cholesterol or diffuse A-beta deposits) are clinically unimportant or cannot be the nidus for devastating later symptoms. Given the enormous complexity of the cellular and biochemical changes that are being steadily identified in AD brain tissue, I find it remarkable that one has been able to obtain as robust a degree of correlation between lesion density (measured at death) and clinical impairment (measured during life) as has been reported in some studies. This "nagging issue" simply cannot obviate the remarkably strong genetic, biochemical, histopathological and animal modeling data that support a causative role for A-beta accumulation in Alzheimer's disease.