The writing was on the wall as early as 35 years ago, when it became clear that people with Down’s syndrome develop AD (Olson and Shaw, 1969), Hardy said. In his famous paper describing the amyloid protein, George Glenner presciently proposed that dementia in Down’s is simply caused by overexpression (Glenner & Wong, 1984).

In Alzheimer’s, the primary deposited protein is a fragment of a gene product, and the mutations found to cause rare Mendelian AD (in APP and the presenilins) increase its production. To find out what’s going on in the vast majority of AD cases, Hardy’s and other labs have screened the genome in sibling pairs of AD disease. They found that AD siblings share their given APP gene variant more than they should by chance. While Hardy has not identified particular high-risk alleles, he speculates they will prove to be the high-expressing ones. It’s not yet clear whether variability in presenilin expression also contributes. In a separate but related vein, recent work has indicated that BACE activity may increase with age.

Diseases marked by tau-only pathology include FTDP 17 as a genetic form plus sporadic versions of essentially the same condition, called supranuclear palsy or corticobasal degeneration. Genetic variability in either tau expression or tau splicing determines risk for these diseases, Hardy suspects, but at present his data can’t differentiate between the two.

Hardy and others noticed that human populations have two tau haplotypes, which differ in intron size. Each is inherited as a block. Curiously, this haplotype is a bizarre structure that, at 2 megabases long, is fifty times larger than haplotype blocks typically are, Hardy said. His group is trying to find out why no recombination would have occurred over such a long stretch of DNA. Homozygosity for the H1 tau haplotype is a risk factor for sporadic tauopathies, as 95 percent of patients have two copies of it, while only 60 percent of the general population do. The H2 haplotype occurs only in people of European ancestry, and both H1 and H2 are more similar to the chimpanzee tau haplotype than to each other. Half in jest, Hardy tossed out an idea that is far-fetched but worth a closer look. Could the H2 haplotype be a genetic legacy of Neanderthals? “A question in anthropology is: Did the Neanderthals kill or breed with contemporary early humans? If history is any guide, they probably did the former with the men and the latter with the women,” Hardy said.

In prion diseases, the primary depositing protein is PrPsc. The first family with a prion mutation was described in the late 80s. In sporadic disease, John Collinge showed that homozygosity at the PrP locus is a predisposing factor, whereas being heterozygous for a common polymorphism confers some resistance. The human PrP locus is unusual in its high degree of heterozygosity; indeed the HLA locus of antigen-presenting genes is one of few other examples known in the human genome, Hardy said. Last year, Collinge suggested that epidemics of kuru-like disease decimated prehistoric populations and in the process created a selective pressure for PrP heterozygosity in human evolution (Mead et al., 2004). In a similar way, evolution is thought to have selected for heterozygosity for the sickle cell gene because it protects against malaria. And yet, while PrP homozygosity increases risk for prion disease, genetic variability at the PrP locus is an even more important factor. The simplest explanation there, too, is that variations in expression levels account for the differences, Hardy said.

Lastly, Hardy discussed diseases with Lewy body pathology, that is Parkinson’s disease (PD) and dementia with Lewy bodies (DLB). Here, too, mutations in α-synuclein can cause familial disease, and genetic variability in the promoter of the normal gene affects one’s risk of developing sporadic PD (Farrer et al., 2001), possibly through higher expression (Chiba-Falek & Nussbaum, 2001). The clearest evidence for the effect of genetic variability on disease risk comes from a family whose early-onset PD traces back to a founder triplication of the α-synuclein gene in the 19th century. This family has no mutation, but members who inherit the affected chromosome end up with 4 copies of the α-synuclein gene instead of the normal 2. “This is the Down’s of Parkinson’s,” Hardy said (see ARF related news story). In Huntington’s disease, too, early indications hint that high-expressing promoters contribute to risk of sporadic disease. Last but not least, genetic variability at the SOD locus should be studied with regard to risk for developing ALS, Hardy added.

Mathias Jucker, who last year joined the new Hertie-Institute for Clinical Brain Research in Tuebingen, Germany, agreed with Hardy that increasing concentrations of proteins prone to misfolding and aggregation are the fundamental commonality between these diseases. He studies the question: once protein concentration has reached a critical concentration, how, then, does aggregation begin?

First, however, Jucker asked where brain amyloid comes from. Prior work had suggested that it enters the brain from the periphery, or that local microglia, smooth muscle cells, or neurons produce it. Jucker favors the latter. He pointed, for example, to analysis of the rare hereditary disease cerebral hemorrhage with amyloidosis-Dutch type, which shows that the amyloid derives from neurons even though most of the deposits build up around blood vessels (see also ARF Live Discussion). To address this question experimentally, his group recently generated a new strain of mice that expresses human APP only in central neurons, not in smooth muscle cells.

This is how Jucker suggested amyloid forms: Neurons release Aβ from the cell body or synapses, interstitial fluid transports it over long distances, and then it enters peri-arterial drainage and gets cleared into blood and the lymphatic system, see also Roy Weller’s presentation. A concentration-dependent, stochastic seeding process along the interstitial pathway explains how deposits form in the vessel wall and in the spaces between neurons. Weller noted that seeding may begin in the vessel wall, and that this may then block the perivascular flow and drainage, resulting in accumulation of Aβ upstream in the parenchyma, particularly while neurons keep releasing Aβ.

To test the idea further, Jucker picked up the concept of seeded polymerization developed by Peter Lansbury, and shown in vivo by Lary Walker. Then at Pfizer, Walker injected AD brain extract into the neocortex of young Tg2576 APP transgenic mice too young to develop plaques on their own. Then he documented the formation of massive amyloid pathology near the injection site and its spread from there (Walker et al., 2002). Jucker extended this work with single injections of diluted brain extract from old APP23 transgenic mice into the brain of their young counterparts. These experiments confirmed Walker’s in that massive amyloidosis formed around the injection site in a dose- and time-dependent way, in which seeding was the rate-limiting step. This happened only in mice that express human Aβ, not in wild type.

Much like prion researchers scratch their heads over precisely what their infectious agent really is, Jucker also asked: What is the seed? Intriguingly, none of many concoctions of synthetic Aβ the scientists injected was able to seed aggregation at the low concentrations that brain extract did. And yet, the seed must be a form of Aβ, because mixing the brain extract with anti-Aβ antibodies prior to injection, or injecting antibody into previously seeded mice both prevented seeding. This suggests that some mysterious feature of endogenous, human Aβ is necessary to induce aggregation. “It may be a folding issue,” Jucker said, “where we cannot reproduce the pathogenic conformation of the native Aβ with synthesis. Or it may be an additional factor that makes Aβ refold in such a way that it becomes a seed, such as a proteoglycan or a chaperone that is present in vivo but not in vitro. Like in prion research, this comes down to structural biology.”

Finally, to jolt awake an audience that was fading at 9:30 pm after 12 talks and a poster session, Jucker asked whether his and Walker’s experiments make AD a transmissible disease, and Aβ an infectious agent. Granted, delivering the brain extract orally did not seed amyloid, and intraperitoneal injection experiments are still ongoing, Jucker said. Even so, the echoes with prion diseases are eerie. There, too, debate revolves around what makes PrPc assume the conformation of PrPsc, and whether this happens in an interaction between PrPc and PrPsc or together with an external factor X? Like in prion disease, amyloid pathology arises from an interaction between a normal and a mutant form of the protein at hand. PrP knockout mice cannot be infected with PrPsc, and even people with familial AD are heterozygous, meaning they have both normal and mutant APP, Jucker said.

Whatever the answer to this tangle of questions, Jucker believes that in sporadic AD, seeding occurs as a stochastic process once Aβ levels rise due to slight increases in production or clearance problems of various kinds. Together, his work also adds indirect support to therapeutic efforts targeted at early-stage amyloid formation, Jucker said.

Intersecting well with Jucker’s work is the hypothesis presented by Roy Weller of University of Southampton School of Medicine, England. Weller proposes that sporadic AD develops as the aging brain gradually loses its ability to clear Aβ along perivascular fluid drainage pathways.

Several mechanisms act in parallel to rid the brain of excess Aβ: Carrier proteins mediate its uptake into blood (see ARF Live Discussion), microglia and astrocytes ingest it, and enzymes such as neprilysin, IDE and others degrade Aβ locally. In fact, Hasan Mojaheri of the University of Zurich, presented experiments extending Takaomi Saido’s original work with neprilysin. Mojaheri told the audience that mice upregulate neprilysin in response to challenge with Aβ, but that the enzyme can only degrade non-aggregated forms of the peptide, and neprilysin induction makes a difference only in young, pre-symptomatic mice.
But back to Weller, who studies a fourth clearance mechanism, namely transport along blood vessel walls. Vascular amyloid usually accompanies AD, and the anatomical pattern of amyloid within artery and capillary walls in the brain suggests that Aβ deposition may begin with its entrapment in the narrow drainage pathways there, Weller hypothesizes. To visualize these pathways, the researchers injected soluble dextran tracers roughly the size of Aβ into adult mouse brain and then colocalized them with laminin to the basement membrane of capillary and artery walls. The resulting pattern of deposition and drainage models Aβ drainage, Weller said.

Cerebrovascular disease induces an age-related congestion of interstitial flow of Ab because arteriosclerosis reduces the amplitude of arterial pulsations, which generate the putative motive force for the perivascular drainage of Ab, Weller said. Evidence for this is that thrombotic occlusion of cortical arteries results in accumulation of Ab in vessel walls upstream of the thrombus, he added. Weller and James Nicoll contributed a synopsis of this hypothesis to a recent on vascular factors in AD, see Alzforum discussion. See also Weller and Nicoll, 2003 and Preston et al., 2003).

How to Study a Folding Protein? Grab It, Watch It.
If indeed the misfolding of an accumulating protein marks a crucible for pathogenesis, how is one to study it? On this question, Harald Janovjak stood in for principal investigator Daniel Mueller, both at the Max-Planck-Institute of Molecular Cell Biology and Genetics in Dresden, Germany, to highlight what the developing field of atomic force microscopy could do for neurodegeneration researchers. The Mueller lab has spent the last few years working out protein imaging and structure analysis methods with the atomic force microscope (AFM), and some may just be ready for prime time.

This desktop instrument offers two general advantages. First, that it allows one to image proteins in relatively native conditions, such as in physiologic solutions or embedded in artificial membranes, and do so at a handsome resolution of up to 0.5 nm. Secondly, it’s possible to physically grab a single folded protein, pull it taut, release it, and measure the tiny, sequential forces with which it resists the unfolding and then snaps back into shape. These force traces then enable the scientist to make inferences about structural characteristics of the protein.

Having a crystal or NMR solution structure of the protein at hand is helpful but not necessary, Janovjak said. In fact, the AFM can predict the outline of a protein’s secondary structure, and such predictions have been confirmed by atomic structures elucidated later on.

Janovjak illustrated some new AFM capabilities. For example, he showed images of gap junctions that visualized individual connexon units. Moreover, the scientists imaged a conformational change as the gap junctions closed in response to an increase in calcium (Muller et al., 2002).

Furthermore, the AFM allows the researcher to image intramolecular forces. For example, consider pull-and-release experiments on bacteriorhodopsin, a complex, well-studied intramembrane protein. The tip of the AFM stylus is attached to a flexible cantilever. Mueller and colleagues retracted the cantilever, measured the force-extension spectra, and correlated these spectra to the extraction of that single protein. This analysis told them that pairs of two transmembrane helices unfold pairwise and in this way act as one structural unit. The German scientists also detected particular secondary structure elements, and studied differences in how they unfold and then refold into their configuration when released (Muller et al., 2002). Finally, Janovjak showed examples of how tinkering with the experimental conditions—temperature, pH, ion concentration—can affect the folding of a given protein (Janovjak et al., 2003).

This ability to modify fairly native conditions experimentally is but one reason why the AFM might lend itself to the study of Alzheimer’s and prion diseases, Janovjak suggested. Another is that many of the proteins of interest are located in membranes. In AD, the role of cholesterol and its metabolites, as well as their possible effect on APP processing, has become a research priority. What’s more, APP proteolysis may well occur in lipid rafts, and their composition is thought to change with age. Could one, for example, insert APP into lipid membranes of different composition and see how such environmental changes affect its folding? The Muller lab invites suggestions for research collaborations in the area of neurodegenerative diseases at (Mueller@mpi-cbg.de).

Retro Traffic Reigns in Protein Renegades?
Weller and Jucker study events occurring after Aβ has left the neurons. But some protein misfolding and oligomerization may well begin inside neurons, see for example Takahashi et al., 2004). How does the neuron try to defuse this ticking time bomb of misfolding and aggregating proteins? This question got attention from Wiep Scheper at the Academic Medical Center in Amsterdam, the Netherlands. Scheper studies Rab6A, a small GTPase that functions in the trafficking of proteins through the ER/Golgi membrane network. Previous work on APP trafficking and presenilin had pointed Scheper’s attention to Rab6, and in St. Moritz she reported data on more direct experiments testing its possible role in AD. Working with Frank Baas and colleagues, Scheper found that, in AD post-mortem brains, Rab6A is upregulated and predominantly expressed in pyramidal neurons of the temporal cortex and hippocampus.

Five years ago, German researchers had found that Rab6 travels along a then-newly discovered retrograde transport route from the Golgi back to the ER (White et al., 1999, Girod et al., 1999). Normally, misfolded proteins are caught in the act in the ER, expelled through the Sec61 channel and fed to the proteasome for degradation. Yet some proteins elude this quality control, and make off into the Golgi on their way to the plasma membrane. Scheper suspects that this Rab6-dependent retro route serves to return such escapees to the ER. She cited prior indications that Rab6 might function in quality control (Luo and Gallwitz, 2003, including of the PrP protein (Beranger et al., 2002). Scheper studies Rab6A in relation to the ER chaperone BiP and reported a close correlation between the two in human brain, possibly because Rab6 induces BiP. This work is ongoing, but Scheper suggested that aging, mutations, or proteasome inhibition may overwhelm the cell’s capacity to retrieve and degrade misfolded proteins. —Gabrielle Strobel