31 Oct 2003

Interview with Dennis Selkoe. The Alzheimer Research Forum posted an interview with Dennis Selkoe of Brigham and Women's Hospital in Boston in 1999. Since then, many new findings have come to light. How has the field shifted? Here is an update.

Q&A

ARF: First, has the primary hypothesis that drives your lab group changed in the past three years?

DS: Not substantially. Any hypothesis must be revised with new data, but it's similar to what it was then. I looked back recently at a paper I wrote in Scientific American in 1991, and the ideas and the artist's cartoon still pertain today. That was before any mutations were known that caused Alzheimer's. There are more similarities to three years ago, 10 years ago, and maybe even earlier, than there are differences.

ARF: What are the major shifts in emphasis that your lab has picked up from the last three years of research across the field?

DS: For our work, one of the major shifts in emphasis we've been following is the interest in small oligomeric assemblies of amyloid-β protein. We knew that such existed in synthetic systems, and we already knew since 1997-98 about ADDLs from Bill Klein and colleagues, and protofibrils that Dominic Walsh, Peter Lansbury, Dave Teplow, and others had described. But it has become increasingly apparent in the last few years that those might actually be the main toxic species. This nuance we didn't know. This theme is reprised in many neurodegenerative diseases, namely that dysfunction seems due less to large aggregates, more to small oligomers, and not to monomers. That wasn't so clear in 1999.

Going back earlier to work in 1988 by Harayasu Yamaguchi, Blas Frangione's group, and others, we knew about diffuse plaques. They were not mature fibrillar amyloid, so the concept that there were deposits in the brain prior to mature amyloid plaques certainly should have given us an idea that something came before the mature plaque, and we stated this in the "amyloid cascade" diagrams that I wrote. The idea of diffuse plaques being upstream of mature plaques has been there since they were discovered. Upstream of diffuse plaques are soluble oligomers that are not visible morphologically in the brain by staining. This development is typical of science: You keep going backwards from what you can see initially toward the origin of the problem. The emphasis on oligomers is an important addition to the amyloid hypothesis, but it doesn't change the fundamental thrust of the hypothesis.

On the other hand, many investigators I admire, like Brad Hyman, Dave Holtzman, and others, continue to show evidence of mature amyloid fibrils causing trouble for neurons and being intimately related to dystrophic axons and dendrites. So it's not so simple. It's typical in any unsettled field that you rush from one side of the boat to the other. These days we like to talk about oligomeric toxicity, but I guess it's a whole spectrum, and even the monomer could ultimately turn out to contribute to toxicity.

ARF: Are there data for that?

DS: Some. For example, the work of Katsuhiko Yanagisawa in Japan shows that some Aβ monomer can be bound 1:1 to GM1 ganglioside (Kakio et al., 2002 and Yanagisawa et al.,1998.) This shifts the monomers from 4 kD to 5 kD on gels. The question is whether that's a form of monomer that has special properties because of the addition of the lipid moiety of the GM1 ganglioside. He has some evidence for that.

For the plain, unmodified monomer, there isn't much evidence that it's toxic, but I guess anything to excess can be a problem. Because these various species are all in equilibrium, we may never really be able to learn exactly what's the contribution of the big polymer, the little polymer, the small oligomer, and the monomer, because the neurons are bathed in all of these. Total Aβ monomer is normally present in your spinal fluid and mine at ~2-4 nanomolar, and in plasma at ~200-800 picomolar. If monomer goes to 10 nanomolar in the brain and CSF, or 20 nanomolar under some conditions locally, it wouldn't surprise me if that yielded trouble from the monomer. It also depends on what's the critical concentration locally in the brain necessary to push it into dimers and trimers. I think that the monomer pool can rise considerably before there's oligomer assembly. I agree that there is no compelling evidence that monomer is toxic, but oligomers, it looks to me that yes, they are.

Our view of the role of oligomers seems a qualitative change. Other things have changed, as well, but I think of those more as quantitative adjustments in terms of our amount of knowledge about something we'd imagined before. For example, the concept of γ-secretase has been around since Aβ secretion was discovered in 1992. It was clear that secreted Aβ must derive from a normal intramembranous cleavage of APP, although it wasn't fully conceptualized as such initially. That was even before that the beautiful work of Brown and Goldstein on the SREBP protein and its intramembrane cleavage. I'd say APP was probably the first protein that was described as undergoing an intramembrane cleavage of some sort, even though no one knew the mechanism. So now, we know a lot about the γ-secretase complex and that's all very interesting, but it is not a departure from the earlier hypothesis; we just know in more detail what the biochemical nature of γ-secretase is.

ARF: You've developed an active interest in synaptic dysfunction as an early step in the development of AD. How do you correlate Aβ oligomers, diffuse deposits, and plaques with clinical progression?

DS: We debate this in the lab all the time. My guess is that by the time the patient has the first noticeable symptoms of memory impairment, even the syndrome of mild cognitive impairment (MCI), she already has a spectrum of species of Aβ in the brain. Of course, early patients have monomers; they've had those since before birth, but they also have small oligomers, they have higher oligomers, they have diffuse plaques, and they have mature fibril-rich plaques. We know this, for example, from the data of John Morris, Joe Price, and colleagues at Washington University, who looked closely at the brains of so-called MCI patients who died without fulfilling clear criteria for AD. These patients already had many classic lesions. So by the time the first clinical symptoms appear, I'm afraid there's already a complex array of Aβ assembly forms.

If you ask what happens 10 years before the first clinically apparent symptoms, or 20 years before, my guess is that monomer levels rise (for many different reasons), and there's an oligomerization, then for quite a while there may just be soluble oligomers that can be seen in the supernatant of the brain homogenate after 100,000 g for two hours, like the oligomers we see in certain conditioned culture media. When there are enough oligomers, they presumably begin to coalesce into larger polymers; you may see this as the first wispy bits of diffuse plaque. But we know from biopsies that by the time the patient has even telltale clinical symptoms of impending trouble, mature plaques are already there.

So, I think it's a continuum. The concept of the oligomers being responsible for the earliest memory change, not the plaques, is complex. In my review (Selkoe, 2002), I discuss the notion that oligomers that are diffusible and soluble are interesting candidates to directly affect synaptic function. We have studied synaptic plasticity paradigms such as long-term potentiation. When we model oligomer toxicity in model systems like a living rat, or in hippocampal slices, we can show synaptic plasticity changes (see ARF news story). (Incidentally, our lab prefers not to work with synthetic oligomers, but uses naturally occurring oligomers produced by certain cultured cells.) But probably, at the moment that oligomers theoretically interfere with synaptic plasticity in MCI patients, larger reservoirs of amyloid material are already around, as well, everything from Aβ polymers, and even fibrils, that are in equilibrium with the soluble forms. In vivo, oligotoxicity at synapses may be happening, but the oligos are in equilibrium with larger aggregates at that time. Whether those larger aggregates are themselves distorting the curvature of axons and dendrites, as Brad Hyman's data suggest for fully developed Alzheimer plaques, I don't know. I can imagine that happening very early, even before the symptoms. Microgliosis, astrocytosis must be occurring, too.

It may, thus, be very hard for us to sort out the precise temporal profile, because so many things happen at once. But even so, I still think that it would be good for synapses in the patients' brains if we could lower the number of oligomers. The synapse may be the place where things go wrong.

ARF: What's the more important research focus today? Synaptic dysfunction or neuronal loss? For years, neuronal loss seemed the most pressing phenomenon in need of an explanation. In Parkinson's, ALS, and Huntington's, too, the question was why are neurons dying. Is that shifting to what's making synapses sick?

DS: Yes, and appropriately so. The synapse likely is the earliest site of dysfunction. It could even be that before synapses become functionally perturbed, microglial cells become activated by oligomers, or astrocytes become activated, and they, of course, contribute to the trouble by releasing cytokines. But at the level of the neuron, I think the synapse is key early on, before neurons die. In the earliest stages of MCI, although plaques and tangles are already there, maybe there's less frank neuronal loss at this point than there is in mild Alzheimer's or in moderate or severe disease. I think neuronal dysfunction causes people to think poorly, and neuron loss likely comes after neuronal dysfunction and contributes to the progressive dementia. I feel that, in general, not only in Alzheimer's but also in Parkinson's, Huntington's, and other degenerative diseases, the focus on frank loss of cell bodies that you can count under the microscope may have been misplaced.

From a therapeutic point of view, I think this issue is not critical. We know what we have to do, regardless. If one did want to answer the question of the contribution of cell loss to very early disease, though, one could do careful stereologic counting of neurons in different regions of hippocampus, entorhinal cortex, and temporal cortex in MCI patients versus age-matched normals and quantify exactly the mean neuronal loss in the earliest MCI brains, assuming you can get a sufficient number of brains of people who died for other reasons during the MCI phase. In short, I think that subtle synaptic dysfunction is likely to be clinically more important early on and precedes synaptic loss, which, in turn, precedes total neuronal loss. Probably, axonal terminals can die back while the cell body is still present.

ARF: Some of the work implicating synaptic loss as an early event—structural and functional decrease of synaptic markers and their correlation with cognitive problems— is really quite old, at least the pathology. What brought synaptic changes to the fore at this time?

DS: A wonderful question. I think it was the maturing of our understanding of amyloid biochemistry to be able to study early assemblies. That includes the recognition of protofibrils and ADDLs as intermediates in fibrillogenesis (even though we still don't know for sure they happen in vivo like they do in the test tube). People developed protocols to study early assembly intermediates. Before that, people did less sophisticated modeling in which they took rather high concentrations of synthetic monomer, made fibrils, spun them out, and characterized them. Importantly, Alex Roher identified SDS-stable oligomers in the human brain as early as 1996 (Roher et al., 1996); others described them a little later, and we reported in 1995 the appearance of soluble oligomers of Aβ in cell culture. I think the field didn't know what to do with the latter observation and perhaps considered it some kind of culture artifact, but we are now convinced that these oligomers are real and form naturally. Between 1995 and 2002, when Dominic Walsh's paper appeared in Nature, we steadily increased our understanding of these highly soluble, low-n oligomers. An important observation that encouraged us in the lab came from Weiming Xia, who showed that both presenilin and APP mutations increased the amounts of the soluble oligomers in our CHO cell culture model, supporting the disease relevance of these oligomers.

Synaptic abnormalities were visualized by electron microscopy all the way back in the 1960s, when investigators like Michael Kidd and Bob Terry performed EM analyses of Alzheimer's brain tissue obtained at autopsy or rarely by biopsy. Bob Terry has been emphasizing for a long time that synaptic loss is the best correlate of the clinical dementia, and many of us interested in the Aβ problem agreed. What we disagreed with was Terry's perspective that the amyloid plaques didn't correlate with dementia, and thus, amyloid was not a critical factor in inducing dementia. In his morphological analysis, plaques indeed didn't, but that led him to believe that amyloid could explain neither the pathogenesis nor the clinical dementia, and he has often said to me that he doesn't favor the idea that amyloid is driving the disease. So, there were early observations of synaptic change, but those favoring a role for amyloid weren't yet able to document that perhaps synaptic alteration was due to very early forms of Aβ accumulation, not visible plaques. Aβ oligomers had not been discovered yet, although any polymer chemist could have told us that when a monomer ultimately forms large polymers, it must go through various intermediary assemblies.

Certainly Peter Lansbury's work on the idea of an Aβ "crystallization seed" that induces fibrillization of the peptide was out there a decade ago (Jarrett et al. 1993; Jarrett and Lansbury, 1993). So the recognition of a synaptotoxic role for Aβ per se—more than the fibrous plaques it produced—came in bits and pieces and, typical of science, it takes a while until things become clearer. I'm sure our viewpoints about precisely how synaptic failure arises in AD will shift further in the months and years ahead.

So, the early morphologists who pointed to important synaptic abnormalities in the pathogenesis of dementia were right. I don't think they were ignored. The question at the time was: "What's the cause of that synaptic change and how does it fit with the emerging genetic data that Aβ could be the overall cause of Alzheimer's? In the last three to four years, it's been possible to get actual hard data that the two fit together. Lennart Mucke's beautiful work looking at synaptic alterations in transgenic mice before any plaques appear, before the earliest diffuse plaques, is very interesting, and there are a number of related studies that I can't quote here due to time constraints, but that I summarized recently [Selkoe, DJ Alzheimer's Disease is a Synaptic Failure" Science, Oct. 2002]. One of the most exciting moments in science occurs when there is a confluence of evidence to support a new viewpoint, and now the accumulated evidence is quite strong. Still, we can't yet conclude today that in the patient with Alzheimer's disease, soluble diffusible Aβ oligomers are the main offender. It probably is a complex mixture.

ARF: In this context, what's your view on the oligomer-specific antibodies reported by Charlie Glabe's lab (Kayed et al., 2003)? They would visualize in vivo what chemically has to be there but was never visible before?

DS: That is the hope.

ARF: What are the implications, and what should these antibodies be used for next?

DS: I found Charlie's paper fascinating. Looking at it closely, sometimes I scratch my head and say, "It's almost too good to be true; it works so perfectly." That he developed an antibody that only recognizes the oligomeric form, not the monomer, not the fibril —that's beautiful. One could imagine such a conformational epitope might occur. That he was able to show that it's shared with oligomers of numerous other amyloidogenic proteins is striking. It's also surprising to my mind that the same concentrations of these different proteins like transthyretin and insulin, prion protein, amylin, synuclein, those identical concentrations of disparate sequences, all led to an oligomeric state that was recognized by the same antibody at the same concentrations. For example, I would have thought that synuclein might have a different critical concentration than Aβ or amylin, but in his graphs, everything seems to come along in sync. In other words, Charlie and his colleagues started with the same protein concentrations in all of the monomer solutions, and they allowed them to oligomerize over the same period of time. It's remarkable that they all behaved so similarly.

As I read the paper, Charlie and his colleagues reported that the cell toxicity induced by these different proteins was closely similar for all of them at the same concentration. I would have predicted otherwise, and find that surprising from a biochemical perspective. But the bottom line in the paper is that they obtained a polyclonal antiserum that recognizes a special conformer of multiple different amyloid proteins, and that the antibody inhibits the toxicity of all of these proteins.

It's hard to imagine that it will work out precisely like that in the human organism, that all amyloidogenic proteins behave similarly, i.e., all at the same concentration and incubation time.

That kind of antibody would be appealing as an immunotherapeutic, because you would apparently do a surgical strike on the oligomer. On the other hand, that the antibody recognizes only oligomers might actually be problematic. It will be beautiful to understand the relative role of Aβ oligomers in synaptotoxicity and behavioral impairment in living animals. I can't wait to see what soluble oligomers do to cognitive performance and to synaptic loss, compared to preparations that include all three major classes: monomer, oligomer, polymer. If I were an Alzheimer's patient, I think I'd just as soon take an antibody that targets all forms. Look at Dale Schenk and colleagues' characterization of what makes an antibody a good clearing tool (see ARF related news story). He suggests that its binding to plaques, its ability to recognize mature amyloid, is more important than its ability to capture soluble Aβ in an ELISA. If that's true, then you might not prefer Charlie's antibody, because he has shown that it doesn't decorate classical amyloid plaques well, though it may decorate what may turn out to be some kind of diffuse material that's quite close to classical fibrillar amyloid.

ARF: You would interrupt the chain of fibrillization...

DS: ...you might. But if you did a surgical strike on the oligomers and there were large amounts of polymer in the patient's brain, and they had an off-rate—which almost certainly they do—and they're turning over, just as some of Brad Hyman's longitudinal data in Down's syndrome or Alzheimer's have suggested, then even if you hit the oligomers, the polymer would release more oligomers over time. Charlie believes his antiserum doesn't see low-n oligomers. He has kindly provided Dominic Walsh with an aliquot, and we'll see whether his antibody detects the low-n soluble oligomers that we observe in cultured cells. At least some of our oligomers are dimers, trimers, and tetramers, and Charlie tells us his antibody wouldn't detect well below an octamer. Also, his work is in synthetic Aβ species, and we're studying natural oligomers that are made and secreted by living cells in very low amounts.

ARF: Should Charlie's antibody be tried in animal models?

DS: Definitely. To see how it does compared to conventional antibodies. The models and protocols for doing that are all in place.

ARF: Has your view of intraneuronal Aβ changed? To name a few developments on this issue, Frank LaFerla's triple-transgenic mice have intracellular Aβ and synaptic LTP effects at six weeks, prior to Aβ deposition (Oddo et al., 2003). Gunnar Gouras has seen Aβ accumulate intraneuronally with immuno-gold EM in human and rodent brain (see Alzforum live discussion). And Andrea LeBlanc found that tiny amounts of Aβ injected into cultured neurons were highly neurotoxic (Zhang et al., 2002).

DS: ... it has. Some of the work that Bruce Yankner did a few years ago in Down's syndrome, showing accumulation of intraneuronal Aβ early on and then less of it visible later on in Down's brains, plus the data you just quoted, suggests to me that we underemphasized the potential importance of intraneuronal Aβ.

At this moment in time, I am more convinced than I was a few years ago that intraneuronal Aβ is likely playing a pathological role. I originally doubted that it was biologically important vis-à-vis neuronal dysfunction because in Down's syndrome, one could see intraneuronal Aβ from the excess expression of AβPP, but even so, the Down's patients fare well for their young and middle years between five and 20. They learn new information and do quite well, and then decline later on in the presence of large amounts of extracellular Aβ, which actually accumulates as early as 10-12 years old. We know from Cindy Lemere's work that by this age, there can be a large number of diffuse plaques, and those are clearly extracellular accumulations (Lemere et al., 1996; Mori et al., 2002). So, the time when Down's patients begin to have detectable dementia-like symptoms (superimposed on their varying degrees of mental retardation from birth) correlates with the appearance of diffuse plaques, then later some neuritic plaques, tangles, microgliosis, and astrocytosis. At the early ages when you see intraneuronal Aβ but no diffuse plaques, the patients appear to be learning, going to sheltered workshops, etc. I focus on Down's because it's really the only naturally occurring human analog of Alzheimer's disease that we know. You can really say with a Down's patient that usually by the age of 40, we will see full-blown Alzheimer's neuropathology and some evidence of clinical dementia. Those who work on intraneuronal Aβ have convinced me that it has a real role to play, but the Down's data still leave doubt in my mind about whether it is the main mediator of neuronal dysfunction.

The bathing of the synapse in soluble, diffusible, oligomeric Aβ I conceptualize as coming from the outside. Of course, intraneuronal Aβ could conceivably be transported down to the synaptic terminal and accumulate there intracellularly. Both intraneuronal and extracellular Aβ are present in AD brain, and we know that Aβ is largely a secreted product. Obviously, all Aβ is initially generated intracellularly, most of it is destined for export, and a lot of it does get out. Which locus is more important may be a question that we can never fully answer in the human brain in vivo. My guess at this moment is that extracellular Aβ accumulation and toxicity are more important pathogenically than intraneuronal. In various models, when we inject into animals, when we treat cultured cells, we always apply Aβ extracellularly and it does cause trouble in many different investigators' hands. For example, when Roger Nitsch's group microinjected Aβ into the P301L tau mutant mouse, they induced a tangle-like process, i.e., tau hyperphosphorylation (see ARF related news story). That gave us what we had been searching for: evidence that Aβ is upstream of enhanced tau phosphorylation and tangle formation, but Nitsch's paradigm was all extracellular. So, I believe intraneuronal Aβ accumulation is there and likely contributes to clinical symptoms, but I guess right now that more of the trouble is coming from outside than inside the neuron.

ARF: Let's talk about inflammation. Your review with Howard Weiner tells me you're thinking extensively about it, and working on it. It's a complex process with probably beneficial and detrimental aspects, so I'd like you to summarize your current view of its role in Alzheimer's.

DS: I'll try. Like intraneuronal Aβ, like the potential importance of oligomers, like synaptic loss, the threads of a role for inflammation were there from very early work, going back at least to the recognition of microgliosis in amyloid plaques and later to the studies of Piet Eikelenboom and colleagues on complement activation around 1982. Much work took off from there, with Joe Rogers, Pat McGeer, and others contributing. Epidemiological studies began associating antiinflammatory drug use with less Alzheimer's, etc. Is it now apparent that this topic is more important than one thought originally? Certainly yes, compared to the early 1980s. But for more than 10 years now, a number of scientists have paid serious attention to the inflammatory mechanism. By the early 1990s, I began including an inflammatory step in my hypothetical schema of the amyloid cascade.

I think one can now assume that microglial cells and astrocytes begin to be bothered by small amounts of aggregated Aβ, possibly already in its oligomeric form. This question hasn't been addressed as clearly for microglia and astrocytes as it has for synapses, but I suspect that soluble aggregates are the initial instigator. The inflammatory process is clearly important, but as with other things we've discussed above, it seems that numerous adverse events happen together, and it's going to be hard to decipher the precise sequence of steps in humans. It could be that Aβ acts through an inflammatory cascade, by which I mean microgliosis-complement activation-astrocytosis-cytokine release, and it could also interfere with synaptic function directly, as we attempt to model when we microinject soluble Aβ oligomers into a living rat. In that experiment, there may be insufficient time for induction of inflammatory mechanisms, because we can see a negative effect on long-term potentiation within minutes.

So, important inflammatory changes occur in AD, and they likely contribute to the development of dementia. One wishes they weren't there, as one assumes that they're bad news. But that's the other question: There is an argument from the work of Tony Wyss-Coray, Lennart Mucke, and others that this inflammation might serve a beneficial role, as well (see ARF related news story; ARF news story). On balance, my sense is that it would be better not to have a lot of microgliosis and astrocytosis.

I would speculate that a purely antiinflammatory approach to the disease would be less effective than an approach that also addresses Aβ economy. Many people believe an eventual combined approach might be best. In this regard, I'm excited about the beautiful discovery of Eddie Koo, Todd Golde, and colleagues that ibuprofen and some other nonselective (COX-1/COX-2) NSAIDs modulate the γ-secretase complex (see ARF related news story; Eriksen et al., 2003; Weggen et al., 2003). They appear to be allosteric, partial inhibitors. They affect the site of cleavage specificity in a way that fits with the model Mike Wolfe and I have put forth for how γ-secretase works, namely, that the presenilin intramembrane aspartates interact with helical substrates, and that cleavage at Aβ42 would be approximately one helical turn above cleavage at Aβ38. That's what Koo's and Golde's work suggests. We had postulated in 1999, when we claimed that presenilin was an intramembrane-cleaving aspartyl protease activated by autoproteolysis and thus represented the γ-secretase, that the aspartates encounter the substrate when it's in (or close to) helical conformation. I think the Koo/Golde data fit beautifully with that model, because ibuprofen doesn't significantly affect Aβ40 cleavage, but you decrease cleavage at 42 while enhancing that at 38, presumably by subtly changing the conformation, and thus substrate interaction, of the two PS aspartates. One goes from a minor amount of 42 to even less 42, and a minor amount of 38 to a bit more 38, while 40 stays about the same. What a cool observation!

ARF: In collaborative work with Howard Weiner, you found T cell clones activated by Aβ in AD patients and normal older people (Monsenego et al., 2003). Does this mean people have autoimmune T cells poised to be stimulated by an immunotherapy? Does this necessitate prescreening for trials?

DS: Alon Monsenego, a postdoctoral fellow in Howard's lab, examined the peripheral blood cells of my patients. I drew blood on a number of my patients and on their spouses (with informed consent), as spouses are often good controls, because they've lived in the same environment many years. A striking finding came out of this. Alon found that people in their thirties, forties, fifties rarely showed peripheral blood mononuclear (PBM) cells that reacted to Aβ when challenged in vitro, but that this was significantly more common in nondemented elderly people and also in patients with Alzheimer's. The prevalence of such peripheral T cells that reacted with Aβ was about the same in Alzheimer's subjects and age-matched controls. So far in this work, it was the aging phenomenon that seemed to lead to much elevated Aβ T cell reactivity. Alon cultured peripheral blood mononuclear (PBM) cells, exposed them to Aβ early on, and then rechallenged them with Aβ over a period of two weeks. He found that there was a small subset of peripheral T cells that reacted to Aβ. This finding was novel, in that there was essentially no information in the literature that Alzheimer's subjects or other humans had intrinsic T cell response to Aβ. There had been some work from Felicia Gaskin and others a decade ago or so on the occurrence of autoantibodies to Aβ in blood (Gaskin et al., 1993), but as regards T cells, only one study had been done, and it reported that there were no T cell responses in Alzheimer's subjects (Trieb et al., 1996). Using a more sensitive assay, the Weiner group has found the opposite: a heightened level of T cell responses in Alzheimer's subjects (as well as elderly controls) compared to middle-aged subjects.

The implications of Alon's findings are that there seems to be a set of peripheral T cells that are primed to react to the autoantigen Aβ and to pour out cytokines, etc. Alon has characterized the type and cytokine profile of these T cells in the paper. It appears that at least some people develop a natural T cell immunity to Aβ. Do these T cells help the process, do they make it worse, or both? Why do people develop these autoreactive T cells with age when immune response in general is declining? Our finding suggests that if you challenge AD patients with exogenous Aβ in a vaccine, such T cells may be primed to expand clonally and spew out proinflammatory cytokines. Had we had these data three years ago, one might have screened individuals for T cell reactivity before enrolling them in the vaccine trial. One may wish to do that in any future active immunization trials. Once others have replicated our finding of autoreactive T cells, one can decide whether this is a screening test that would be useful. Even if one pursues the vaccination approach Howard Weiner, Cindy Lemere, Ruth Maron and I developed, i.e., mucosal (oral or nasal) Aβ treatment, one might wish to screen for intrinsic T cell reactivity. It's a new concept and raises the question of whether T cell reactivity contributes to the development and/or progression of Alzheimer's disease.

ARF: Why do Aβ monomer levels rise in people without FAD mutations? Do we need to know?

DS: Yes, we do. I suspect there are a number of distinct reasons that older humans develop an imbalance between Aβ production and Aβ clearance. These may include defects in degrading proteases or Aβ transport mechanisms. For example, ApoE4 inheritance may well represent a type of clearance failure.

ARF: On the question of Aβ degradation, where can drug developers dig in? Boosting an enzyme's function, say neprilysin or IDE, is generally less practical that blocking it.

DS: This is true, but that doesn't mean it can't be done.

ARF: You published a paper with Dominic Walsh and others on what happens to the intracellular domains (ICDs) of APP family members once they've been cleaved by γ-secretase (Walsh et al., 2003) They either get degraded or are stabilized by binding to Fe65 proteins, in which case, they might affect gene expression. The signaling consequences of these ICDs are a hot topic these days. Can you venture a guess on what sorts of genes and biological processes they likely control? And will this thread lead us, finally, to a physiological function for APP, APLP1, and APLP2?

DS: I favor the hypothesis that the APP proteins are receptors that release their signaling cytoplasmic domains via regulated intramembrane proteolysis. In our lab, Taylor Kimberly and Bing Zheng have developed nice evidence for AICD binding to Fe65 and nuclear localization in both non-neural cells and primary neurons. This is, indeed, an exciting topic that should ultimately clarify the normal function of APP after all these years. The original cloning of full-length APP by Beyreuther, Masters, Mueller-Hill, and colleagues in 1987 suggested that the molecule was a receptor, and I think this will turn out to be right.

ARF: One area your lab is not pursuing is oxidative stress, yet a PubMed search on the term and Alzheimer's pulls up 690 citations. Where do you fit oxidative damage into your view of the disease? What's the specific link to AD?

DS: It's an interesting part of the AD cascade, and I believe it occurs downstream of gradual Aβ buildup in limbic and association cortices, and the initiation of synaptic and microglial injury by soluble oligomers. Because our lab's focus is on events that we think are very early and thus potentially more amenable to prevention (and because there are only 24 hours in the day), we have not chosen to investigate oxidative stress, but I have no doubt that it plays an important role in the insidious development of neuronal dysfunction

Gabrielle Strobel Interviews Dennis Selkoe

Quick Links

Tools

Dennis Selkoe

Comments

Make a Comment

References

Other Citations

External Citations

Further Reading

Annotate