The talks at this meeting fell into these categories:

β-syn, Lentivirus
GDNF
Immunotherapy and CAA update
Secretase inhibitor update
NSAID update
Statin update
β-sheet breakers

Eliezer Masliah of the University of California, San Diego, discussed the interplay of syn family members in human patients and transgenic mouse models. β-syn is expressed at even higher levels than α-syn in the brain, and the mRNA levels of syn family members differ from controls in AD (relatively more γ-syn) and LBD (relatively more α-syn). β-syn prevented α-syn aggregation in vitro even under oxidative conditions, and lentiviral delivery of β-syn decreased α-syn aggregation in cell culture. Moreover, β-syn suppressed α-syn aggregation in vivo: Crossbreeding with β-syn transgenic mice and lentiviral delivery of β-syn both reduced protein aggregation in α-syn transgenic mice, and synapse loss was ameliorated. Conversely, AD pathology appeared to exacerbate α-synopathy in α-syn transgenic mice: Crossbreeding with AβPP transgenic mice that produce excessive Aβ42 exacerbated α-syn aggregation and phenotype, possibly involving oxidative stress conferred by LRP-mediated Aβ42 uptake. Interestingly, Aβ40 suppressed α-syn aggregation. (see ARF related news story).

Patrick Aebischer, Lausanne University Medical School, Switzerland, gave a superb overview of the potential for lentiviral gene delivery systems to generate animal models and therapeutic approaches for neurodegenerative diseases. First, dominant genes (α-syn, Htt, SOD1, etc.) could be silenced by application of siRNA-coding lentivirus (see ARF live discussion). Proof of principle was achieved for viral siRNA knockdown of GFP in cell lines and silencing of SOD1 in a transgenic mouse model of ALS. Second, expression of recessive genes (parkin, DJ-1, etc.) could be restored. Third, neuroprotective and neurotrophic genes could be delivered. For example, lenti-GDNF injected into the substantia nigra of MPTP-intoxicated primates caused impressive behavioural improvement, total rescue of 18fluoro-DOPA PET, and fiber sprouting. Aebischer noted that it might even become possible to satisfy individual demands of therapeutic doses of GDNF by the use of tetracycline-regulatable lentiviral constructs.

Deniz Kirik of Lund University, Sweden, explained that acute dopamine neurotoxin models are extremely effective, but too rapid for useful neuroprotection studies. In a milder rat model based on intracranial injection of low-dose 6-hydroxydopamine, the neurotrophic factor GDNF was found to be neuroprotective when delivered repeatedly to terminals during the degenerative process, it even protected the rats against locomotor deterioration. First, clinical trials showed that intrastriatal infusion of GDNF alleviated symptoms in five Parkinson’s patients to an extent similar to standard L-DOPA monotherapy (see ARF related news story). Kirik went on to describe the phenotypes of rat and monkey models of Parkinson’s based on adeno-associated viral gene delivery of α-syn, which recapitulate early α-synopathy and reversible, selective dopamine neuron loss (see ARF related news story).

Dale Schenk of Elan Pharmaceuticals, South San Francisco, California, reviewed approaches to Aβ immunotherapy in three categories: active immunization with Aβ, active immunization with immunoconjugates (Aβ fragments conjugated to an ovalbumin-derived T cell antigen), and passive immunization with purified monoclonal antibodies against Aβ.

All three approaches reduce Aβ burden in transgenic mouse models. The first caused a dramatic reduction of plaque burden and gliosis in two different transgenic mouse models, as well as cognitive improvement. Cross-reactivity with soluble Aβ was observed, but capture of soluble Aβ did not correlate with therapeutic efficacy.

Optimizing Aβ immunotherapy requires knowing the most effective epitopes and immunoglobulin class. AN 1792 immunization yields antibodies against epitopes mostly in the N-terminus of Aβ in mice and humans. IgG2a (2C1 and 12B4) immunoglobulins proved the most effective subtype for clearing Aβ and reducing the neuritic burden in passively immunized mice, probably because this antibody subtype interacts most efficiently with Fc receptors (see ARF related news story).

Schenk also discussed two potential side effects of Aβ immunotherapy. Transgenic mice with a high burden of vascular amyloid suffer cerebral hemorrhage after passive immunization with a monoclonal IgG1 directed against an N-terminal Aβ epitope (see Mathias Jucker’s talk). T cell invasion into brains of C57Bl/6 mice was observed upon AN 1792 immunization plus concomitant pertussis toxin treatment (see ARF related news story). Both efficacy and side effects were investigated in the first autopsy study of an AN 1792-treated patient (see ARF related news story). In this patient, Aβ immunotherapy was effective as evidenced by large plaque-free areas in the temporal cortex. Aβ-phagocytosing microglia were visible. This patient suffered excessive inflammation in the brain that was recognized in an MRI scan and controlled by dexamethasone treatment, though she never fully recovered. She died of a pulmonary embolism 12 months after the last dose. Despite the marked reduction of senile plaques in this immunized patient, neurofibrillary tangles and cerebrovascular amyloid persisted in the disease course of this person.

Christoph Hock, University of Zurich, continued the theme by reporting on the Zurich cohort (30 Aricept-treated AD patients) of the immunization trial (see ARF related news story). Immunized patients developed strong and sustained anti-Aβ immune sera. The human antisera intensely and selectively stained compact, diffuse, vascular plaques but displayed no cross-reactivity with soluble Aβ. Such antibodies were also collected from some immunized patients’ CSF. Interestingly, the antibody titers did not correlate with staining intensity, suggesting distinct avidity of the antibodies in each patient. Antibody titers also did not correlate with the incidence of aseptic meningoencephalitis, which could be treated with methylprednisolone. Publication of the first report on cognitive amelioration of the immunized AD patients can be expected by the end of this year.

Mathias Jucker, University of Basel, Switzerland, explained his hypothesis of the mechanism of vascular amyloidosis derived from transgenic mouse models. Expression of AβPP exclusively in neurons of (Thy1)-AβPP mice bred into an AβPP null background causes cerebral amyloid angiopathy to develop, demonstrating that neuronal expression of AβPP is sufficient. Blood levels of Aβ are low in this mouse model, so vascular amyloid is unlikely to be blood-borne. An AβPP dutch mouse model of hereditary cerebral hemorrhage with amyloidosis reveals that all the amyloid is in the cortical vasculature. Upon aging, these mice have hemorrhages and neuroinflammation. AβPPdutch mice show very little 42-ending Dutch Aβ, but this can be induced upon crossbreeding with PS1G384A. In conclusion, high Aβ42 levels lead to plaque formation in the neuropil, whereas high Aβ40 levels allow amyloid transport along the perivascular drainage pathway, where vascular amyloid eventually builds up.

In AβPP23 mice, passive immunization with β1 monoclonal IgG1 against Aβ (3-6) mainly decreases diffuse plaques and Aβ42 levels. Immunotherapy has no beneficial effect on CAA; indeed, these mice have significant cerebral hemorrhage, possibly because of the enormous burden of vascular amyloid in this particular mouse model (see ARF related news story). Under these circumstances, decreases in cerebral blood flow were evident by functional MRI on anesthetized mice. Finally, Jucker cautioned that although EM analysis reveals microglia in close proximity to amyloid, Aβ was never visualized inside microglia. Moreover, the "sink hypothesis" of decreased Aβ levels leading to plaque removal is still unproven. Thus, the mechanism of plaque clearance upon Aβ immunotherapy remains to be solved (see also ARF related news story).

Martin Citron of Amgen Inc, Thousand Oaks, California, talked about BACE1, the enzyme conferring β-secretase activity. Because β-secretase cleavage is the first step in Aβ production, and BACE is elevated in AD cortex (see ARF related news story), it is a prime target for antiamyloidogenic drug development. Peptidomimetic BACE1 inhibitors are published, but small-molecule BACE inhibitor development is difficult. Early attempts with inhibitors developed against proteases with sequence similarity (e.g., HIV protease) were disappointing. Solving the crystal structure of BACE1 revealed a highly complex active center with no less than eight subdomains, possibly inspiring rational drug design.

What side effects might be expected from therapeutic BACE inhibition? In addition to AβPP, there are additional substrates (ST6gal I, PSGL-1, and probably more) that could be vitally dependent on BACE cleavage. Nevertheless, knocking out BACE1 in mice caused no phenotype, even upon aging. Even though these mice have neither β-secretase nor Aβ, they have no pathology and no changes in gene expression as measured by Affimetrix mouse chips. Proof of concept for an antiamyloidogenic effect of BACE1 inhibition was provided by the finding that crossing Tg2576 with BACE1-/- mice suppressed plaque formation in aged offspring.

Mark Shearman, Merck Sharp and Dohme Research Laboratories, Harlow, Essex, United Kingdom, described a novel series of small-molecule γ-secretase inhibitors. WO 0236555 is the lead compound of sulfonamido-substituted, bridged bicycloalkyl derivatives that decrease plasma Aβ, soluble brain Aβ, and AβPP CTFs. Screening for selective AβPPγ (42) cleavage inhibition provided evidence that certain NSAIDS (ibuprofen, flurbiprofen, sulindac, etc.) act as noncompetitive γ-secretase inhibitors. These compounds also inhibit Notch cleavage at high concentrations. In fact, 60 compounds compared for AβPP cleavage and Notch cleavage revealed no differential efficacy at all. This cross-reactivity of γ-secretase inhibitors points to the potential side effect of altered immune responses due to blocked Notch-dependent hematopoiesis (i.e., T cell development).

Edward Koo, University of California, San Diego, reported that ibuprofen lowers Aβ42 production independent of cyclo-oxigenase (COX) inhibition, because the Aβ42-lowering properties are maintained in fibroblasts from COX-/- mice. Not all NSAIDs reduce Aβ42. Some, for example, celecoxib, actually increase Aβ42 with a reciprocal reduction of Aβ38, while others, for example, sulindac, lower Aβ42 production concomitant with an increase of Aβ38. In fact, among the NSAIDs clinically assessed for AD to date, indomethacin was the only one that ameliorated AD symptoms, and it was also the only one with Aβ-lowering properties. It will be interesting to optimize drugs for a maximal Aβ42-lowering effect and eliminated COX cross-reactivity. A two-center phase I trial with R-flurbiprofen in 48 healthy elderly (55-80y) subjects just started, assessing safety, tolerability, pharmacokinetics, and blood- and CSF-Aβ biomarker effects in AD patients.

The molecular mechanism of the Aβ42-lowering effect of some NSAIDs is not clear. AID/AICD generation or stabilization by Fe65 is apparently not decreased, arguing against direct γ-secretase inhibition. However, in CHO cells cotransfected with FAD presenilin mutants, NSAIDs generally have a stronger Aβ42-lowering effect (with the notable exception of _ex9 presenilin-1). Perhaps certain NSAIDs exert allosteric γ-secretase modulation. (see related NSAID coverage; NSAID live chat).

Mika Simons of the University of Tübingen, Germany, mentioned that cholesterol depletion leads to an 80-90 percent reduction of Aβ production, while adding back cholesterol restores Aβ levels. This is due to a cholesterol-mediated decrease of α-secretase and increase of β-secretase activity. The reason for this effect may be correlated to the amount of AβPP in rafts, where β-secretase cleavage might occur. In brain there is no cholesterol uptake from LDL, rather it must be synthesized de novo. Thus, BBB-permeant statins like simvastatin could selectively lower intracellular cholesterol levels in neurons. Indeed, treatment of guinea pigs with simvastatin lowers brain Aβ levels with minimal side effects (e.g., changes in liver enzyme levels); all effects are reversible upon simvastatin washout in the guinea pigs.

Retrospective metastudies indicate that hypertensive patients treated with statins had a lower prevalence of AD. Prospective studies of statin treatment against coronary diseases and stroke with cognitive monitoring were inconclusive, so AD-specific studies are warranted. The first clinical trial of simvastatin in mild-moderate AD patients proved somewhat efficacious (Simons et al., 2002 http://www.alzforum.org/pap/annotation.asp?powID=20097). Plasma LDL cholesterol decreased by 50 percent, CSF free cholesterol decreased by 10 percent. MMSE scores improved slightly in mild cases only, and this correlated with decreased cholesterol levels. No beneficial effect was observed in moderately impaired AD patients, perhaps the plaque load may have already been too high. CSF tau levels were not decreased.

Claudio Soto at Serono Pharmaceutical Research Institute in Plan-les-Ouates, Switzerland, first summarized general issues of protein misfolding disorders. They include intracellular vs. extracellular inclusions, peptide-specific vs. generalized inclusions, the deposition of misfolded protein aggregates vs. misfolding of soluble (dysfunctional) protein, loss-of-function vs. gain-of-toxic-function, and the involvement of posttranslational modifications.

Soto talked about developing β-sheet breakers based on peptides that intercalate with the amyloid-seeding peptide sequence of aggregation-prone proteins spiked with conformation-breaking proline residue(s). In the case of Aβ, such an anti-aggregation peptide specifically prevents Aβ (but not prion or amylin) fibrillization in vitro as well as in AβPPld-transgenic mice. Administration of the β-sheet breaker ameliorated the astrogliosis and microglial activation, as well as the minor neurodegeneration normally observed in this mouse model. β-Sheet breakers led to improvement in water maze performance in a rat model of amyloidosis caused by Aβ injection, but there was only partial reduction of plaque burden. The groups’ current focus is on the development of Aβ-sheet breakers with improved pharmacokinetics. Soto said that the toxicology so far is encouraging and no immunological responses to β-sheet breaking peptides were seen. Compound optimization strategies include the development of small peptidomimetics and nonhydrolysable peptide derivatives. Peptidomimetic approaches are being pursued for various different indications but have not, to date, generated novel drugs, as technical challenges remain.

Leon Thal, also at University of California, San Diego, concluded the program by reviewing the clinical course of AD with a specific emphasis on the design of prospective clinical trials. He pointed out that the study end-points must be carefully selected and followed through. More importantly, patients must be recruited early in the disease, ideally right around the first manifestation of mild cognitive impairment, or even prior to symptoms. He also noted that although AD varies considerably from patient to patient, the progression of this disease is rather uniform and predictable, even in limited cohort sizes.

This was a great meeting that brought together interesting people who presented excellent science with great enthusiasm. Never has it been clearer that improved therapy is within reach for Alzheimer’s disease, and that improvements are underway for Parkinson’s, as well. The enchanting scenery of the locale served as an allegory for the hopeful vista we received toward the future.—Philipp Kahle, Ludwig Maximilians University of Munich, Germany, and Bart De Strooper, Flanders Interuniversity Institute for Biotechnology and KU Leuven, Belgium.