The talks at this meeting fell into these categories:
Gamma-secretase complex and BACE
Other RIP proteases
AD/PD genetics
Lipoprotein receptor signaling
Tau function
Tau, α-syn, Ab aggregation

With the inaugural talk, Dennis Selkoe of Brigham and Women’s Hospital in Boston, Massachusetts, set the standards by discussing mainly unpublished material. The first part of his talk was about the reconstitution of γ-secretase activity in mammalian cells (see Kimberly et al., 2003). Overexpression of nicastrin, presenilin-1, Aph1 and Pen2 are necessary and likely sufficient to restore γ-secretase activity as measured in a cell-free assay. Coimmunoprecipitations, copurification on an affinity column with immobilized γ-secretase inhibitor, and comigration in glycerol gradients all suggest that the four proteins are in one active complex. A multistep fractionation procedure that involves extraction with one percent DDM and a purification step pulling down FLAG-tagged γ-secretase components after detergent change to one percent digitonin results in sufficiently pure material that can be analyzed in silver-stained SDS-PAGE. The members of the complex are prominently present, but some additional bands are difficult to interpret (contamination or really members of the complex?). Mass spectrometric analysis of these proteins is ongoing.

Dennis Selkoe also elaborated on the AICD (the famous amyloid precursor protein (AβPP) intracellular domain believed to be involved in signaling). This fragment translocates to the nucleus in association with Fe65, and its generation is developmentally regulated in primary neuronal cultures differentiating in vitro. Maximal AICD production is observed when the cultures start to generate synapses. The production inversely correlates with phosphorylation of the intracellular domain of AβPP, specifically Thr-688. Delta and Jagged, two Notch ligands, can be added to the expanding list of γ-secretase substrates (see ARF related news story). Quite impressive evidence was discussed demonstrating that the liberated Jagged intracellular fragment can reach the nucleus to stimulate the signal from an AP1 reporter assay. The ADAM17/TACE-cleaved, membrane-bound Jagged fragment competes with Notch for cleavage (providing a negative feedback on Notch signaling). Further work on the Aβ degradation activity of insulin-degrading enzyme (IDE) was shown. Most importantly, IDE-/- mice display decreased Aβ degradation, resulting in 50 percent increase in steady-state levels of Aβ in the brain (see ARF related news story). Interestingly, in these mice, AICD also accumulates (mainly the nonphosphorylated form). Aβ oligomers injected into living anesthetized rats inhibit long-term potentiation in vivo and also in hippocampal slices, and c-Jun N-terminal kinase inhibitors can reverse this effect. In regard to the vaccination studies (see also end of the meeting), it should be noted that the meningoencephalitis-causing T cell response is mainly induced by epitopes in the C-terminal parts of Aβ, while the N-terminal parts are involved in the generation of the (hopefully curative) humoral response.

Finally, a word on Parkinson’s: an O-glycosylated (N-acetylglucosamine and sialic acid-containing) species of α-syn called α-Sp22 has been identified in human brain (see ARF related news story). This α-Sp22 binds to parkin, but not any longer when deglycosylated. α-Sp22 has been purified from brain and was found to be glycosylated on Ser-9 or Thr-22. This raises the question of how this cytoplasmic protein gets glycosylated.

Stefan Lichtenthaler of the Ludwig Maximilians University of Munich, Germany, presented a novel substrate for β-secretase (BACE). He checked different candidate proteins (L-selectin, TNFR2, AβPP and PSGL-1) and found that, of this series, only AβPP and PSGL-1 "shedding" could be increased upon BACE transfection. In BACE1-deficient cells, this processing is extinguished; transfecting BACE restores it. PSGL-1 is P-selectin glycoprotein ligand-1 and mediates leukocyte rolling on the endothelium, allowing transmigration and tissue invasion of leukocytes. Cleavage of PSGL-1 occurs in the juxtamembrane domain.

Patrick Keller of the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden, Germany, reviewed the many indications for a role of cholesterol in AβPP processing. He went on to discuss the specific association of BACE and a fraction of AβPP with "lipid rafts" (Ehehalt et al. 2003). These are glycosphingolipid subdomains in the cell membrane. YFP-[wt]AβPP and BACE-CFP co-patch with placental alkaline phosphatase, but not with transferrin receptor[?5-41] in such cell surface rafts. Antibody-mediated cross-linking of the rafts increases Aβ production in a cholesterol-dependent manner, probably by fusing rafts and bringing more AβPP and BACE in contact with each other. Some evidence suggests that BACE and its cell surface substrates (such as AβPP) come together during endocytosis by "raft clustering," increasing Aβ production and secretion.

Takeshi Iwatsubo, University of Tokyo, wondered about the individual roles of nicastrin, APH-1, and PEN-2 in the presenilin complex. RNAi knockdown of each individual protein abolished Aβ production in Schneider cells and γ-secretase activity (both with AβPP and Notch substrates) in membrane preparations. Drosophila PEN-2 (dPEN-1) RNAi causes an accumulation of full-length Drosophila presenilin that is stabilized and engaged in a high molecular weight complex. However, co-RNAi of dAPH-1 or dNicastrin abolishes this full-length presenilin accumulation. These effects were also seen in HeLa cells, SY5Y cells, Neuro2a cells and mouse primary neurons. Overexpression of dPEN-2 facilitates the formation of a high-molecular-weight presenilin complex by increasing the accumulation of presenilin fragments, and results in increased γ-secretase activity. Human PEN-2 can rescue RNAi-mediated dPEN-2 deficiency concomitant with conversion of full-length presenilin to fragments. [G112D]dAPH-1 (conserved Gly in TM4) mutation abolishes the stable expression of dNicastrin and dPEN-2 and, consequently, γ-secretase assembly. Overexpression of the three cofactors with presenilin increases the accumulation of active presenilin fragments and γ-secretase activity. These observations lead Iwatsubo to propose the following model: Presenilin is synthesized as light-molecular-weight species, then assembled in a stable high-molecular-weight complex by binding of APH-1 and nicastrin, which stabilize the holoprotein. Presenilin is finally activated upon endoproteolysis facilitated by PEN-2. PEN-2 itself does not resemble a protease, though this possibility is not entirely excluded. It is, however, more likely that PEN-2 acts as an accessory factor, or that it brings presenilin into a cleavable conformation. Some of this work just appeared in Nature (see Takasugi et al., 2003).

Harald Steiner, Ludwig Maximilians University of Munich, reported that nicastrin RNAi prevents formation of the γ-secretase complex. In stable nicastrin RNAi cells, wild-type nicastrin rescues nicastrin maturation, presenilin fragmentation, PEN-2 and APH-1 stabilization, and γ-secretase activity. Mutational analysis indicates that the whole ectodomain of nicastrin is important. Limited trypsin proteolysis suggests a conformational change of nicastrin upon γ-secretase complex formation (immature nicastrin is trypsin-labile; mature nicastrin is trypsin-resistant). This is not due to glycosylation, because SDS unfolding allows trypsin degradation of mature nicastrin, and prevention of glycosylation by kifunensine does not influence trypsin sensitivity. Thus, conformational compaction of nicastrin accompanies its maturation (see Shirotani et al. 2003).

Steiner pointed out that PEN-2 is an essential component of the high-molecular-weight γ-secretase complex. PEN-2 RNAi prevents the formation of the γ-secretase complex and also decreases nicastrin maturation. Each component of the γ-secretase complex is dependent on the expression of all others.

Steiner also reported reconstitution of the γ-secretase complex in yeast, an organism that does not have endogenous γ-secretase. Coexpression of all four components in yeast is associated with presenilin endoproteolysis and allows the liberation of an AβPP-based reporter construct from the membrane. The presenilin D385A active site mutant abrogates this activity. Thus, coexpression of presenilin, nicastrin, APH-1, and PEN-2 is sufficient for γ-secretase activity; no other components are needed, Steiner said. He reported coimmunoprecipitation experiments confirming that the four components are in contact with each other. The reconstituted presenilin-1 complex has γ-secretase activity; C100 is cleaved into Aβ and AICD, and cleavage produces Aβ38, Aβ40, and Aβ42, as well as AICD50 and AICD51. This work just appeared in Nature Cell Biology (see Edbauer et al., 2003).

Bart De Strooper pointed out that the molecular mass of all four components of the γ-secretase complex is about 220kD, and this molecular mass complex is indeed observed in transfected cells. However, the native complex is twice that size (440kDa), raising the question whether dimeric complexes exist. Could there be multiple presenilin complexes that might have differential, perhaps organ-specific functions? De Strooper noted that systematic Northern analysis revealed that liver and kidney have particularly high expression of all γ-secretase components. Aspartate mutants of PS1 are entirely inactive, although they are accurately incorporated into the presenilin complex (including PEN-2 stabilization and nicastrin maturation), as evidenced by complementation of PS1-/- PS2-/- cells.

Nicastrin glycosylation is presenilin-1 gene dose-dependent, but not needed for γ-secretase activity, as kifunensine prevents nicastrin maturation, but nicastrin still makes it to the cell surface and allows formation of γ-secretase activity (see Herreman et al., 2003). PS1+/- PS2-/- mice show numerous benign skin tumors, a condition similar to human seborrhoic keratosis that is common in the elderly. Tumors also form when double-knockout mice are rescued with Thy1-driven presenilin-1 (Xia et al., 2001). Loss of presenilin alleles increases the concentration of β-catenin and phospho-β-catenin in keratinocytes and fibroblasts, pointing to the importance of presenilin in Wnt signalling in the skin (Kang et al. 2002). Moreover, adult mice have a severe autoimmune phenotype, including enlarged spleens and salivary glands. There are immunoglobulin deposits in several tissues, and autoimmunoglobulin reactions are detectable in the sera of these mice, but the epitope of auto-Ig is not known. Nevertheless, this is a potentially dangerous side effect that must be monitored in γ-secretase inhibitor trials.

Ralf Baumeister, Ludwig Maximilians University of Munich, reported that the C. elegans presenilin-1 homolog Sel-12 is expressed throughout the worm’s life cycle, while the presenilin-2 homolog hop-1 gradually increases expression in larval stages. Spe-4, a third homolog, is expressed only in the L4 stage. Sel-12 mutant worms have an egg-laying deficit due to defects of the vulvar smooth musculature. Possibly, cerebral bleeding in PS1-/- embryos is caused by a similar defect of blood vessel musculature. Loss of temperature memory indicates a nervous system defect. The [C60S]sel-12 mutation (corresponding to the Italian [C92S] presenilin-1 mutation in humans) mediates increased Aβ42 production in mammalian cells. Both the egg-laying deficit and the neuronal phenotype can be rescued by human presenilin-1 wild-type and only partially by presenilin-containing clinical mutations. In general, FAD mutations are loss-of-function mutations in C. elegans. Therefore, Baumeister suggested that γ-secretase inhibitors that reduce activity even further could increase the problems of AD. He proposed instead to increase presenilin activity to restore Aβ40 production.

A suppressor screen for sel-12 rescue yielded five candidates that rescue the egg-laying deficit and protruding vulva phenotype with higher than 90 percent penetrance. None of the suppressors work in sel-12/hop-1 double mutants; thus, all suppressors depend on hop-1 expression. Spr-5 strongly derepresses hop-1 expression. Spr-3 and spr-4 encode REST-like zinc finger transcription factors. Spr-5 is homologous to polyamine oxidases, which together with CoREST (spr-1) associate with histone acetylases and with the REST transcription factor, all of which act together in chromatin remodelling. The bottom line is that this mechanism derepresses hop-1 expression, allowing for replacement of sel-12 with hop-1. Does hop-1 take sel-12’s place in the complex? (See also Lakowski et al. 2003.)

Matthew Freeman, of the MRC Laboratory of Molecular Biology, Cambridge, United Kingdom, reviewed the rhomboids, a novel family of intramembrane-cleaving (i.e., RIP) proteases. Spitz, a TGFα homolog in Drosophila is cleaved by Rhomboid, a multipass transmembrane protein inside the Golgi membrane. Star is needed for transport of Spitz from the ER, and without Star, cleavage of Spitz does not occur. Rhomboid is an intramembrane serine protease distinct from the aspartyl proteases presenilin and signal peptide peptidase. It works as a single enzyme and not as catalytic subunit of a complex. No precleavage of Spitz is necessary for Rhomboid proteolysis. Interestingly, and in contrast with γ-secretase signaling, the extracellular domain released in the Golgi is secreted into the extracellular milieu and serves as the signaling moiety.

Rhomboids are widespread in the animal kingdom. Seven homologues are known in Drosophila, four in mammals, and two in yeast (Rbd1 and Rbd2). What could be the function of rhomboid in yeast, given that this organism does not depend on intercellular communication? To answer this question, Freeman’s lab made yeast knockout strains, and while the Rbd2-knockout is fine, the Rbd1-less strain grows more slowly, especially on glycerol. This points to a respiratory problem. Tbd1 appears to be a mitochondrial protein, and these organelles had a collapsed morphology in the knockout strain. What could the substrate be? In a clever in silico approach, Freeman looked for candidate proteins that met these criteria: mitochondrial, single transmembrane domain, and biochemically characterized as soluble proteins. Candidates included Ccp1p, which is involved in oxidative stress, but its knockout has no consequences, and Mgm1p, a dynamin-like GTPase regulating membrane remodeling. Remarkably, the Mgm1 knockout yields a similar phenotype to the Rbd1 knockout. Further work confirmed that Rbd1p cleaves Mgm1, and mutation of the Mgm1p cleavage site causes the same phenotype (mitochondrial collapse) as the complete deletion of the protein.

Rhomboid is a highly specific protease, cleaving Spitz, but not EGFR, Delta, or TGN38. Even the human homolog of Spitz (TGFα) is not cleaved because it lacks the proper cleavage sequence in the transmembrane domain. Spitz/TGFα swapping revealed that a GA motif around amino acid 140 accounts for this remarkable sequence specificity of rhomboid. The transmembrane helical ASIASGAMCAL sequence is characterized by small and helix-destabilizing amino acids. This sequence was found (manually!) in 20-30 candidates in the mouse genome. Incidentally, the micronemal (i.e., infection-promoting) adhesion proteins in Toxoplasma gondii have similar transmembrane domains and are also subject to intramembrane cleavage. By analogy, rhomboid was found to cleave malaria-relevant MIC proteins. Both findings could yield therapeutically relevant approaches. Thus, control of intramembrane proteolytic cleavage becomes a major issue in many diseases other than AD, as well.

Bruno Martoglio, Swiss Federal Institute of Technology, Zurich, presented an overview of the biology of signal peptide peptidase (SPPase), a presenilin-related intramembrane protein-cleaving protease (see ARF related news story). SPPase is responsible for the generation of the HLA-E epitopes that are generated by cleavage of the signal peptides of certain type II proteins. In the cytosol, a proteolytic activity that remains unidentified (or at least was not discussed at the meeting) chops up the cytosolic fragment. The question arose whether a similar process could be responsible for trimming of the Aβ peptide after release.

Overall, there are many differences between presenilin and SPPase, which is a member of the presenilin homolog family. SPPase has an opposite orientation, is not incorporated in a multimembrane complex, is not activated by presenilinase, and recognizes type II proteins as substrates. Nevertheless, several inhibitors that inhibit presenilin also inhibit SPPase. For instance Merck’s L685,458 inhibitor binds not only to PS but also to SPPase in cross-linking studies. The question arose why this was not noticed in the original publications (see also Weihofen et al., 2003).

Steve Younkin of the Mayo Clinic in Jacksonville, Florida, elaborated on the four candidate AD genes on chromosome 10: PLAU, VR22, IDE, and CYP46. He discussed systematically the evidence in favor of each gene and made a case for pooling the results from genetic studies to demonstrate the significance of the association with late-onset AD. For the four genes, evidence was brought forward for association, even though such association could not be confirmed in John Hardy’s series. This initiated a philosophical discussion on "whether the pint is half-full or half-empty," a problem that was more deeply plumbed in the bar session ("more work is needed"). In any event, PLAU (urokinase-type plasminogen activator)-deficient mice demonstrate an age-dependent increase of plasma Aβ levels. VR22 (α-T-catenin) variants are significantly associated with plasma Aβ42 and the VR22 4360/4783 appears to account for a substantial proportion of chromosome 10 linkage (see ARF related news story). Finally, Aβ is increased in IDE-/- mice and IDE variants are significantly associated with plasma Aβ42 in some studies. Cholesterol-24-hydroxylase (CYP46) significantly associates with AD in a large multicenter study of almost 600 AD vs. 600 controls (see ARF related news story).

John Hardy of the National Institute on Aging in Bethesda, Maryland, mentioned that the chromosome 10 locus appears to contain multiple AD genes. Tested candidates include IDE, PLAU, and VP22, but all turned out negative in three series (Cardiff, Washington University, Mayo Clinic Jacksonville) of more than 300 cases and 300 controls. Apparently, there is a difference of opinion between the Younkin and the Hardy consortium. The prior discussion on the state of the pint came up again, and Hardy concluded the topic with a typical one-liner: "What’s confirmed by John Hardy is fact."

As for Parkinson’s genes, pseudo-dominant inheritance of the parkin phenotype was observed in no less than 10 heterozygote mutations that segregate with disease, pointing to haploinsufficiency. Two SNPs in the parkin promoter segregate with disease and affect promoter activity; these are especially frequent in single-mutation parkin patients (these could be the cases with Lewy bodies). α-syn haplotypes are associated with Parkinson’s. α-syn overexpression, especially its mutant forms, inhibits proteasomal activity in M17 cells and enhances toxicity of proteasome inhibitors, while wild-type parkin overexpression ameliorates this. Upon transfection, a quite specific toxicity of A53T α-syn to tyrosine hydroxylase-positive neurons in ventral mesencephalic neuron cultures is observed.

Finally, the PARK7 locus was rigorously linked to the DJ-1 gene by the identification of about 10 deletion and nonsense mutations (see ARF related news story). When asked what he thought about the involvement of DJ-1 in the cellular handling of oxidative stress, Hardy drew laughter with his second deadpan reply: "Oxidative stress is like apple pie."

Joachim Herz, University of Texas Southwestern Medical Center, Dallas, discussed the mechanisms of signaling by the lipoprotein receptors: LDLR, VLDLR, MEGF7, ApoER2, LRP, LRP1B and megalin. ApoE receptors mediate signal transduction in neurons. ApoE receptors (VLDLR and ApoER2) control brain development. Reelin usually binds to these two receptors and stimulates tyrosine phosphorylation of the adaptor protein disabled (Dab1), activating a cascade from PI3K to Akt to GSK3β to tau hyperphosphorylation (see ARF related news story). Reelin mice, as well as VLDLR- and ApoER2-knockout mice, show increased tau phosphorylation. Dab1-/- mice do, also, but the extent of this is dependent on the genetic background. For example, 129xC57BL6 is permissive, while little or no tau phosphorylation is apparent in the Balb/c. The reelin mutation is viable in this background, as well. Microsatellite analysis of F1 intercrosses of the two strains reveals quantitative trait loci (QTLs). Interestingly, peaks of association were found on chromosome 1, 12, and a strong one on chromosome 16; PS1 and AβPP reside in the QTLs on chromosome 16 and chromosome 12, respectively.

Finally, Herz provided nice insights into novel functions of the ApoE receptors in signaling. PDGF-BB induces tyrosine phosphorylation of LRP via caveolar PDGF receptors, allowing binding of the adaptor protein SHC. This is prevented by ApoE lipoproteins and by the tyrosine kinase inhibitor Gleevec, now used in cancer therapy. Smooth-muscle cell-specific LRP-knockout mice on a high cholesterol diet show stunning atherosclerosis and aortic aneurysms. A Gleevec-laced diet rescues this phenotype. LRP-deficient aortas show an enormous increase in PDGF-R signaling. Thus, this pathway could contribute importantly to the atherosclerotic phenotype. EM Mandelkow

Eva-Maria Mandelkow, of the Max-Planck Institute for Structural and Molecular Biology in Hamburg, provided interesting information on the pathway of how tau stabilizes microtubules and mediates axonal transport. In AD, tau disassembles from microtubules (MTs) and forms paired helical filaments (PHFs), but it is unclear whether these are directly toxic to the neuron. It is possible that MT-bound tau can be harmful, too. Proline-directed phosphorylation yields epitopes that are recognized by the most commonly used AD-diagnostic antibodies, but these phosphorylations barely affect tau binding to MTs. MT disassembly occurs after phosphorylation of the C-terminal KXGS motifs catalyzed by MARK kinases, for example. In fact, MARK2 induces neurite outgrowth, and dominant-negative forms of MARK2 inhibit neurite outgrowth in N2A cells (Biernat et al., 2002). Interestingly, MARK2 tagged with a HA epitope co-localizes with F-actin, and phospho-tau (12E8) localizes to phalloidin-positive actin filaments. Physiologically, KXGS site phosphorylation allows for dynamic microtubules as exist in growth cones, for example.

What, then, is the effect of increased binding of tau to MT? Apparently, tau inhibits plus-end directed transport by competing with kinesin for the same binding site on MT. When tau is overexpressed, more of it binds to MTs, increasingly hindering both antero- and retrograde transport. Mandelkow measured this by determining the run length and velocity of individual vesicles. The velocity does not change, but the run length is shortened in both directions by tau. Since tau also interferes with the binding of kinesin, but not dynein, to MT, its net effect is that retrograde transport becomes dominant. MARK influences this process by causing the removal of the tau obstacle. In conclusion, tau inhibits and MARK facilitates the transport of active mitochondria, influencing synaptic energy production. Tau40 overexpression also inhibits the transport of AβPP vesicles. Tau thus leads to accumulation in the cell body of axonal transport cargoes (synaptic vesicles, mitochondria, etc.) with imaginable adverse effects including, for example, increased sensitivity to H2O2.

Continuing along similar lines, Eckhard Mandelkow stated that tau is a natively unfolded and highly soluble protein, whose fibril formation accelerates in the presence of additional factors such as polyanions (heparin, polyglutamine, DNA). Indeed, phosphorylation prevents aggregation, and it is therefore surprising that PHF can form despite tau being phosphorylated. The VQIVYK sequence in the tau protein tends to form β-structures, and it associates with this motif in other tau proteins to nucleate further assembly, leading to classical amyloid. The FTDP-17 mutations have little effect on MT stability. Instead, P301L and _K280 FTDP-17 mutations in tau accelerate PHF assembly, perhaps by favoring β-strand conformation Tryptophan mutation scans and autofluorescence measurements performed to analyze the vicinity of domains within PHFs reveal that the hexapeptide motifs that form the aggregation nucleus are buried in the PHFs. Interestingly, PHFs in vitro have low intrinsic stability; they are easily denatured by guanidine as monitored by tryptophan fluorescence. Screening of PHF inhibitor compounds is based on thioflavin S fluorescence. Actually, tau bound to MT induces thioflavin S fluorescence like in PHFs, and overloading of tau on MTs leads to some filament-like structures on the surface of microtubules. (See also Barghorn et al., 2002.)