Speakers include:

Harald Janovjak
Eckhard Mandelkow
Judith Frydman
James Bowie
Charlie Glabe

See the movie

What new techniques have come on line to study protein misfolding? Harald Janovjak introduced the audience to the latest refinements of atomic force microscopy, which make this desktop structural biology instrument ripe for attempts to image membrane proteins in neurodegenerative disease. Janovjak, a Ph.D. student at the University of Technology in Dresden, Germany, recapped his talk from a recent conference in Switzerland, and added further detail and perspective. For example, to illustrate how it is possible to correlate polypeptide sequence elements to unfolding, he ran a movie that showed the consecutive unfolding of pairs of helices in a single bacteriorhodopsin molecule pulled by the AFM tip.

Janovjak also talked about the AFM’s ability to detect assembly of protein subunits into functional complexes, a topic of interest in investigations of γ-secretase and APP. The AFM can contribute a variety of measurements to the understanding of protein misfolding, Janovjak said. For example, it can quantify rates of spontaneous misfolding, or probe parameters such as energy landscape, stability, and interaction with environmental components in different conformations of a given protein.

Eckhard Mandelkow, at Hamburg’s Max-Planck Institute in Germany, posed the problem of how one defines misfolding in a protein that is natively unfolded? To date, more than 100 proteins are known to fall into this category, including tau, Aβ and α-synuclein. A natively unfolded protein does not assume an ordered structure. Instead, it is flexible and becomes ordered only upon binding, i.e., when it performs a function, or indeed when it aggregates.

In the Aβ field, the argument regularly comes up that if only the physiological function of Aβ were known, its role in AD would become clear. Ironically, tau demonstrates that this is not necessarily the case, because its function is understood but its role in AD is not. Mandelkow said that tau’s role in axonal transport holds the clues to what it does in AD. Too much tau covers up the microtubules, blocking the movement of vesicles and organelles along microtubules, while too little tau destabilizes the microtubules (see ARF related news story). The balance of tau at microtubules is influenced by the concentration of soluble tau in the neuron, which, in turn, is one of the factors that determine its aggregation. Even so, how aggregation follows from excess soluble tau, and how this relates to axonal transport changes in AD remain a mystery.

One piece of the puzzle may lie in MARK kinases, which regulate tau levels on microtubules (see ARF related news story and ARF news story). Several different kinases can phosphorylate tau at different sites, but to date no consensus exists on whether increased phosphorylation promotes the formation of tau filaments or not. Mandelkow then reviewed research establishing a working model for how tau aggregates, and his lab’s recent establishment of a screen for aggregation inhibitors. Ongoing work will test these inhibitors in a new, inducible transgenic mouse model of tau.

Judith Frydman, of Stanford University, talked about the basic mechanisms of protein folding in live cells. The chemist and Nobel Laureate Christian Anfinson believed that a protein finds its normal folding state by itself. This may be true for some proteins, Frydman said, but in general, folding is more complicated, especially in cells. Two abundant chaperone systems aid the proper folding of newly translated proteins. Hsp70 is a monomeric protein that exists in many homologs in different cellular organelles, and chaperonins are oligomeric, ring-shaped complexes of which the major one in eukaryotes is called TRiC/CCT.

To find out how newly translated proteins fold, Frydman’s group performed dilution experiments similar to what Anfinson had done in his time. In the test tube, their model protein (the firefly luciferase, which cannot fold by itself) took a half-time of eight minutes to fold when placed in a mammalian cell extract containing chaperones. However, in live cells, where the chaperonins wrap around the nascent luciferase protein right as it peels off the ribosome, luciferase folds four times as quickly and with different intermediate states. This suggests that the cell contains a structured chaperone network at the ribosomes, Frydman said.

In further studies of protein folding during translation, this time with the tumor suppressor von Hippel Lindau (VHL) protein, the VHL substrate required the cooperation of Hsp70 and chaperonin to fold properly. Its cancer-causing mutations allow Hsp70 binding but interfere with subsequent chaperonin binding. Indeed, a proteomic analysis of chaperone-mediated folding suggests that a large fraction of cellular proteins transit through chaperone systems in the course of their folding, using first one, then another.

Besides supporting the folding of new proteins, chaperones also rescue proteins denatured by stress. Is the latter just a variation of the former, whereby either a new protein or a stressed one can be the substrate for chaperones but the process is the same? This is still an open question, but Frydman’s current data suggest that the process is a different one. The eukaryotic cytosol, she believes, contains separate chaperone networks that handle protein synthesis and quality control.

Finally, Frydman addressed the question of exactly what chaperones might do with misfolded proteins. They could be passive, merely holding on to the protein until it refolds properly. Or they could be more active, even interacting with ubiquitin ligase to feed a protein that is beyond repair into the proteasome. “We know very little about this,” Frydman said. To approach this issue, her coworkers took natural VHL mutations that generate misfolded VHL protein. They expressed these mutants in different yeast strains, one that degrades the mutant and one that can’t. They found that the chaperonin TRiC, which folds de novo VHL inside its cylindrical chamber, does not participate in its degradation. VHL degradation did require Hsp90, a chaperone not involved in its de novo folding. Other chaperones assist in both folding and degradation. This suggests that folding of new proteins and quality control of damaged proteins fall to distinct, specialized sets of chaperones, Frydman concluded. For more, see Spiess et al., 2004.

Two back-to-back presentations, one on computational structure prediction and one on cell biology/electrophysiology, highlighted a curious new discrepancy about which structure the Aβ peptide might assume. As a structural biologist, James Bowie of the University of California, Los Angeles, has not previously worked on neurodegeneration but ran smack into AD research this year. It happened when Bowie’s group tested a deceptively simple algorithm they had developed to model how certain membrane proteins fold. This research grew out of a frustrated acknowledgement that, despite 40 years of trying, scientists still have not cracked the old problem of de novo structure prediction.

Far fewer membrane proteins than cytosolic proteins have yielded insight into their atomic structure, largely because membrane protein biochemistry is so complex. Bench science aside, however, mere computational prediction of how a protein will fold ought to be easier for membrane proteins than for proteins in watery solutions, Bowie said. That’s because in water, a protein can assume a huge number of conformations, and sorting through them to find the right one remains daunting. With a membrane protein, its sequence implies where its transmembrane regions are, and once scientists have packed these, the possible structures for arranging the rest of the polypeptide chain become greatly restricted, Bowie said.

In this spirit, his group developed a three-step algorithm for folding homo-oligomeric transmembrane helices. Starting with helices in random orientation, the program runs 200 iterations of a procedure that finds conformations of minimal energy determined by the protein’s many van der Waals forces. Next, the algorithm filters out asymmetric conformations, then it clusters similar structures together. The largest cluster contains the structure most likely to be the correct one. “If someone had told me this approach a while ago I would have dismissed it as too simple. But as it turns out, this works really well,” Bowie said.

Bowie presented examples of validated or predicted structures, for example for glycophorin, for an influenza virus proton channel, and for the H. pylorum cytotoxin VacA. Like many of the proteins in this study, VacA appears to form a pore, and the commonality between these proteins turned out to be a packing motif Bowie called a glycine zipper. “We think this is an important mode of creating channel structures,” he said. One in four membrane proteins contains at least one glycine zipper motif—too many to examine. A more stringent database search, for proteins that contain a single transmembrane helix and at least four glycine zippers in a row, dredged up Aβ, Bowie said, as well as major prion protein precursor, which is also thought to form channels.

The idea that misfolded pathogenic proteins form pores in neuronal membranes has been around in the Alzheimer’s, Parkinson’s, Huntington’s, and prion fields for a decade without garnering wide support (see Kawahara et al., 1997, Lin et al., 1997, Kagan et al., 2002, Lashuel et al, 2003, Lashuel et al., 2002, Hirakura et al., 2000, and Kourie and Henry, 2002). Bowie said that his lab first reproduced some of the earliest data reported on the topic (Arispe and Rojas, 1993). Next, scientists began testing their own hypothesis that the glycine zipper motifs drive formation of the purported channels. They mutated different glycine positions on Aβ, and initial results suggest that wild-type Aβ forms channels, but the mutants do not, Bowie said. The mutants also appear less toxic to cultured neurons, but that work is even more preliminary. Bowie cautioned that he has no evidence as yet on whether his work is relevant to Alzheimer disease, and invited the field to come up with ways of finding out.

Ironically, Bowie’s research opens up a fresh vein of support for the channel hypothesis just as prominent work in the AD field appears to weaken it. This website has reported extensively on Charlie Glabe’s and Rakez Kayed’s studies of an antibody that is specific to small oligomers of different amyloidogenic peptides regardless of their amino acid sequence (see ARF related news story). This surprising study had suggested that the antibody recognizes a shared structural motif on the peculiar oligomeric intermediates of these proteins. At the NBA conference, Glabe, who is at the University of California, Irvine, recapped published data and noted that the list of amyloidogenic proteins known to react with the antibody has since grown to 24.

This September, the scientists further reported that these intermediates all damage synthetic lipid bilayers by a common mechanism that increases the membrane conductance, but that the oligomers do not form pores or specific ion channels in the process (see ARF related news story). At the main Society for Neuroscience conference, Erene Mina, a graduate student in Glabe’s lab (who contributes occasional news summaries to Alzforum), presented a poster describing how treating cultured cells with Aβ oligomers greatly increases calcium influx and disrupts the integrity of the membrane. Soluble oligomers of other amyloidogenic proteins do this, as well, but their respective monomers or fibrils do not. The anti-oligomer antibody reverses the change in conductance. However, much as the scientists had expected to find oligomer pores, they could not. “We see no evidence for discrete conductivity, we see no evidence for open and closed states, and we see no ion specificity. We looked a broad range of inhibitors described as Aβ channel inhibitors, to no effect,” Glabe said at the NBA symposium. “I wanted to see a channel, but this is what we are left with. It’s an urgent issue for us to sort out, and we’ll do it soon.” Other recent work also supports the notion that amyloidogenic oligomers damage the lipid bilayer but not by forming channels (see Green et al., 2004.)

In related news, Glabe also reported these data: immunocytochemistry experiments performed to see whether these oligomers exist in human brain found no immunoreactivity in controls, but a punctate pattern in AD brain. The anti-oligomer antibody detected only a small fraction of total amyloid and did not colocalize with astrocytes or microglia. It occurred largely in extracellular regions that contained plaques but did not overlap with plaques or with diffuse Aβ deposits. Rather, it stained the outer rim of diffuse deposits. The oligomer antibody did not colocalize with neurons, though some intraneuronal staining was apparent. Having solved technical problems with Western blots using this antibody, the lab has now detected the oligomers in soluble extracts of people with MCI and AD, but not controls, Glabe added.

Glabe’s laboratory has begun collaborations with groups studying other diseases. One, with Jeffrey Robbins at Cincinnati Children’s Hospital, shows that human tissue from conditions not previously thought to be amyloid aggregation diseases actually show massive staining with the anti-oligomer antibody. Glabe mentioned forms of idiopathic dilated cardiomyopathy as examples, where oligomeric intermediates might exert their toxicity early and formal fibril aggregation never fully progresses (see Sanbe et al., 2004). The antibody may identify further diseases involving amyloid, Glabe said.

In short, Glabe proffered this working hypothesis: The primary mechanism of all degenerative amyloid diseases lies in the oligomer-induced leakiness of membranes, possibly because oligomers disrupt the way membrane lipids are packed (see also Glabe, 2004).—Gabrielle Strobel.

To be continued Monday, 20 December 2004.