Steven G. Younkin, MD, University of Pennsylvania, PhD, University of Pennsylvania - US grants

The 4 kD (39-43 residue) amyloid beta protein (AB), which is deposited as amyloid in Alzheimer's disease (AD), is encoded as an internal peptide that begins 99 residues from the carboxyl terminus of a set of 695-770 residue glycoproteins referred to as the amyloid Beta protein precursor (BAPP). Recent data from our laboratory and others has established that normal processing of the BAPP involves (i) a constitutive secretory pathway in which the BAPP is cleaved within AB to produce a large, secreted, NH2-terminal derivative and an 8.7 kD COOH-terminal fragment, neither of which can produce amyloid because they do not contain the entire AB, (ii) endosomal/lysosomal processing which produces a complex set of COOH-terminal derivatives that includes potentially amyloidogenic forms with the entire AB a or near their NH2 terminus, and (iii) the production of 4 kD AB(essentially identical to the AB deposited as amyloid in AD) that is released from cultured cells and readily detected in CSF. Strong evidence that amyloid deposition plays a critical role in the development of AD has come from the identification of familial AD (FAD) kindreds in which the AD phenotype cosegregates with mutations in the BAPP gene. Three of these mutations alter the valine mutation (NL) alters the lysine-methionine located immediately amino to AB1 (lys670=met671 in BAPP) to isoleucine, phenylalanine, or glycine. A fourth double mutation (NL) alters the lysine-methionine located immediately amino to A1 (lys670-met671 in BAPP770) to asparagine-leucine. The location of these mutations in close proximity to AB immediately suggests that they may cause AD by altering BAPP processing in a way that is amyloidogenic. To evaluate this possibility, we have recently compare human neuroblastoma (M17) cells expressing normal BAPP695 or FAD-linked mutant BAPP 695. Cells expressing the BAPP NL mutant showed a 5-fold increase in the relative amount of an 11.4 kD AB-bearing carboxyl- terminal BAPP derivative, and they released 6-fold more 5 kD AB into the medium. These observations provide strong evidence (i) that this mutant BPP causes AD because it undergoes altered processing that releases increased amounts of AB, and (ii) that the pathway producing AB in cultured cells is highly relevant to AD. In this application, we propose to expand our analysis of AB production by undertaking experimentation designed t o (1) test the hypothesis that the 4kD AB produced and released by normal cellular processing consists of several peptides with variable COOH-termini (AB39-43) similar to the AB peptides that have been identified in AD amyloid, (2) test the hypothesis that the BAPP717 mutants favor production of the longer AB1-42 or 43 forms, which selectivity deposit as plaque core amyloid, without increasing the production of AB as expected from the observation by our group and others that phorbol esters substantially increase the production of the large secreted BAPP derivative. In addition, we propose (4) to compare BAPP processing and the AB released by transfected mouse or human cells expressing wild type BAPP or the FAD-linked mutant BAPPs in order to (i) determine if the disappointing results obtained to date in transgenic mouse model of AD are related to low rates of AB production in mouse cells expressing human BAPPs, and (ii) identify constructs likely to significantly increase CNS AB production when they are introduced into transgenic mice. Finally, we plan (5) to carry out a systematic comparison of the amount of soluble AB in various regions of AD and control brains in an effort to determine if the amount of soluble AB is correlated with AD and/or amyloid deposition.

Our working hypothesis is that over 50 percent of typical late onset Alzheimer's disease (AD) is linked to genetic determinants that increase the concentration of amyloid beta protein (Abeta42) in a way that can be detected in plasma. Our data indicate that these determinants, which do note the coding region of the genes linked to early onset familial AD, cause plasma Abeta42 to be elevated in approximately 34 percent of first-degree AD relatives. We find that plasma A642 increases with aging in control subjects over the age of 65, that plasma Abeta42 is not elevated in most symptomatic patients with typical late onset AD, and that plasma Abeta decreases in the Tg2576 model of AD coincident with cerebral Abeta deposition. Thus we postulate that the typical profile of many AD patients is one in which plasma Abeta42 is elevated or high normal, increases further with aging, and then declines toward or into the normal range as Abeta deposition begins and the symptoms of AD develop. Since declining plasma Abeta appears to be linked to Abeta deposition in brain and Abeta42 is deposited in the brain of everyone who develops AD, plasma Abeta42 may decline in everyone who develops AD, even those whose plasma Abeta42 is in the normal range. This means that the simple, annual measurement of plasma Abeta42 may be an excellent way to identify many of those who are destined to develop AD, much as measurements of plasma lipoproteins and cholesterol identify those at risk for atherosclerotic heart disease. We plan to examine this possibility by following 500 first-degree AD relatives, 400 over age 65 and 100 between the ages of 45 and 65. Plasma Abeta and cognitive performance will be analyzed annually to determine (1) how plasma Abeta changes with aging in this cohort of AD relatives and (2) if elevated and/or declining plasma Abeta are important risk factors for typical late onset AD or memory decline.

DESCRIPTION: The identification of mutations in the APP, PS1, and PS2 genes that cause early-onset familial Alzheimer's disease (AD), the demonstration that these mutations all increase Abeta42, and the discovery of an association between Apolipoprotein E4 and late-onset Alzheimer's disease have dramatically improved our understanding of Alzheimer's disease. It is clear, however, that much of the genetic risk in late onset Alzheimer's disease remains unexplained. Current strategies to identify other genes that affect late-onset Alzheimer's disease have met with limited success often because of the difficulty associated with obtaining late-onset families with sufficient power for reliable linkage analysis. Genetic studies using large numbers of small families or sib-pairs, to increase the power of the analysis, are also currently being performed by several groups however difficulties with the non-replication of positive loci, identified by different studies, has continued. It will also be difficult to identify the biologically relevant genetic variability using loci identified by this type of small family/sib pair analysis, as candidate regions tend to be large and poorly defined. In this proposal we describe how high plasma ABeta levels can be used as a surrogate phenotype to increase the power of late-onset AD families allowing not only the reliable linkage of a specific chromosomal region with disease but also facilitating the identification of the genetic variability that is responsible for the increased plasma Abeta42 and for the increased risk of developing Alzheimer's disease.

The effort to identify genes with alleles that influence susceptibility to late onset AD (LOAD) proceeds logically from linkage to association and then to identification of specific susceptibility alleles. Using plasma amyloid beta protein (Abeta42) levels as an intermediate, quantitative phenotype for LOAD, we obtained linkage at approximately 80 centimorgans (cM) on chromosome 10 (Ch10). Linkage to the same region was obtained independently in a study of affected LOAD sib-pairs. Pursuing these findings, we have now identified three Ch10 genes (VR22, PLAU, and IDE) with variants that show strong association with LOAD and/or plasma Abeta42. The proposed study focuses exclusively on IDE. There is compelling biological evidence from IDE knockout mice that the insulin degrading enzyme is normally involved in Abeta degradation, so IDE is clearly an excellent candidate gene. In our MCJ series, LOAD showed highly significant (p<10[-7]) association with the same IDE haplotypes that the Brookes group showed to be significantly (p<0.01) or highly significantly (p<10[-9]) associated with AD in multiple case-control series from Scotland and Sweden. These same haplotypes showed significant association with plasma Abeta42 in our extended LOAD series. The goal of this study is to identify the LOAD susceptibility alleles in the IDE region. In all LOAD patients from the MCJ series, we will thoroughly screen the coding and putative regulatory regions of the IDE gene and, if necessary, adjacent genes. The SNPs that we find (approximately 60 are expected from our initial screen of IDE) will be analyzed using a two stage statistical approach to identify the set of putative susceptibility alleles that show strongest, most reproducible association with LOAD and with plasma Abeta42. Strongly associating SNPs will then be evaluated for biological effects relevant to LOAD. To facilitate analysis of variants in putative regulatory regions and to increase the size of our case-control cohort, we will prepare cDNA and genomic DNA from the large number of frozen LOAD brains available through the Neuropathology Core.

DESCRIPTION (provided by applicant): There is a growing consensus that the best way to manage Alzheimer's Disease (AD) will be through preventive therapy. To facilitate preventive therapy, it is important to develop AD-related biomarkers that can be used to identify at risk individuals in the same way that cholesterol levels are used to identify those at risk for atherosclerotic heart disease. For this reason, we proposed in the last cycle to determine if plasma AB40 and/or AB42 might be useful biomarkers for identifying at risk individuals. In 563 normal subjects that we followed longitudinally, the plasma AB42/40 ratio was an excellent biomarker for identifying those who developed Mild Cognitive Impairment or AD in three to five years. The cumulative incidence of AD/MCI was significantly greater in subjects with an AB42/AB40 ratio in the lowest quartile as compared to those with a ratio in the highest quartile after adjusting for age and ApoE4. Subjects with an ApoE4 allele and a low (below median) AB42/40 ratio, began to develop AD/MCI at 2-3 years and, by 5 years, over 20% of the subjects in this group were affected. In contrast, only 3% of the ApoE 4 carriers with a high (above median) AB42/AB40 ratio developed AD in five years. Combining age and the AB42/AB40 ratio was also highly effective in separating subjects who developed disease from those who did not. Older subjects (age >80 years) with a low (below median) AB42/40 ratio began to develop AD/MCI at 2-3 years and, by 5 years, over 20% of the subjects in this group were affected. In contrast, less than 4% of all other subjects developed AD within five years. If these findings can be confirmed, it seems likely that the plasma AB42/AB40 ratio can become an important biomarker for developing and implementing a preventive approach to AD therapy. Our specific aims are to (1) confirm that the plasma AB42/AB40 ratio is a useful biomarker for identifying those who will develop MCI or AD in three to five years, and (2) determine if elevated AB (AB40 and/or AB42) is useful for identifying those who will develop MCI or AD in five to fifteen years. Several additional biomarkers will be evaluated in the same longitudinal series where plasma AB is analyzed. Dr. Wyss-Coray will analyze BDNF, AcrpSO (aka adiponectin), angiogenin, PDGF-BB, and MCP-1. Dr. Jack will analyze hippocampal atrophy as well as whole brain atrophy using the Boundary Shift Integral (BSI) approach. The utility of these additional biomarkers will be evaluated singly as compared to plasma Aft and jointly with plasma AB.

[unreadable] DESCRIPTION (provided by applicant): In an effort to understand and to overcome the factors that thwart replication of genetic association studies, we recently studied variants in the conserved regions of the gene (IDE) that encodes the insulin/AB degrading enzyme gene in considerable detail. A remarkably high percentage of these variants had modest effects that showed replicable association when analyzed in our large case/control series. Based on these results, our current working hypothesis is that progress in identifying novel LOAD genes has been slow because most LOAD genes are like IDE; they have multiple susceptibility alleles with modest effect size. The effect of genes with powerful variants like the ApoE 4 allele is easily detected and replicated in small case/control series. But the net effect of genes with multiple susceptibility alleles that have modest effects, though substantial, cannot be detected and replicated well unless large case/control series are employed to evaluate a set of variants selected for their likely functional effect. Using the scientific infrastructure developed in the last cycle we propose to pursue this hypothesis by targeting genes in the AB processing pathway. We will perform an unbiased, genome-wide search for AB QTLs likely to harbor novel LOAD genes in the AB processing pathway. In addition, we will search thoroughly for and examine the function of additional susceptibility alleles in the known, major genes of the AB processing pathway. Our specific aims are to (1) perform whole genome scans to identify novel quantitative trait loci (QTLs) linked to plasma AB levels, (2) identify novel LOAD susceptibility alleles by using multiple, large case control series to analyze the variants in conserved regions of major genes in the AB processing pathway (SORL1, APP, IDE, MME, ECE1, PLAU, BACE1, PSEN1, PSEN2, and VR22), and (3) evaluate the functional effects of the susceptibility alleles identified in specific aim 2. Depending on the specific location of each susceptibility allele identified, function will be analyzed by evaluating the effect of the variant on (i) plasma AB and/or (ii) brain mRNA. Our recent results suggest that many susceptibility alleles may act by altering gene expression or splicing. [unreadable] [unreadable] There is strong evidence that reducing the AB42 peptide in normal elderly subjects could prevent Alzheimer's disease (AD), a disorder that inflicts enormous suffering and financial loss on our society. To perform affordable prevention trials and administer drugs to normal elderly people with an acceptable risk/benefit ratio, methods must be developed for identifying those elderly individuals who are at increased risk for AD. In this application, we propose experimentation to identify many genes with variants that alter AB42 thereby influencing risk for AD; we do so because each new AD gene identified opens new therapeutic possibilities and improves our ability to identify the at risk population. [unreadable] [unreadable] [unreadable]

[unreadable] DESCRIPTION (provided by applicant): Vasopeptidases are enzymes responsible for the generation or inactivation of vasoactive peptides. As such, vasopeptidase inhibition represents a significant approach for the treatment of cardiovascular disease. For example, angiotensin-converting enzyme (ACE) inhibitors are FDA approved and widely used to treat hypertension and heart failure. In addition, inhibitors of two other vasopeptidases, neprilysin (NEP) and endothelin-converting enzyme (ECE), are currently in preclinical and clinical development as are combined inhibitors of these enzymes. Recently, considerable data has emerged indicating a role for the vasopeptidases ECE and NEP, and perhaps also ACE, in the degradation of the Alzheimer's amyloid beta-peptide (ABeta). As the abnormal accumulation of ABeta plays a pivotal role in AD pathogenesis, pharmacological inhibition of these vasopeptidases is of considerable concern as they may increase ABeta levels in a manner that increases risk of developing Alzheimer's disease. In this application we propose to directly evaluate the hypothesis that chronic reductions in ECE, NEP, and ACE will result in enhanced ABeta accumulation and AD-like pathology by examining the effects of mice genetically deficient in these enzymes and mice treated with clinically relevant vasopeptidase inhibitors. The results of these studies will further our understanding of the role of NEP, ECE, and ACE in determining ABeta concentration in the brain, while helping to assess the potential risk of the clinical use of vasopeptidase inhibitors. [unreadable] [unreadable]

DESCRIPTION (provided by applicant): An invariant feature of the pathological cascade in Alzheimer's diseases (AD) is a reactive gliosis, reflecting an underlying alteration in the innate immune activation state within the brain. Innate immune signaling is altered early in AD, but is also skewed towards an activated state as a consequence of brain aging. There is strong genetic evidence that innate immunity has a significant role in AD. Variants in two genetic loci that play roles in the complement cascade, CR1 and CLU, show significant genetic associations with AD, and rare coding variants in TREM2 also confer substantial risk for AD. Numerous experimental studies in AD mouse models show that manipulating innate immune pathways can have positive or negative effects on proteostasis, cognition and neurodegeneration. At least when assessing A¿ pathology as an endpoint, the beneficial effects of some innate immune system manipulations are robust. We propose to identify therapeutic targets within the innate immune signaling cascade in AD that could be safely manipulated to provide disease modification in AD. However, because of the complexity of, and the gaps in our knowledge regarding, innate immune signaling within the CNS, a systems level approach that integrates multiple types of data will be required to achieve this goal. Indeed, development of any innate immune therapy will need to be finely tuned and extensively validated in order to be further developed as a potential AD therapy. We will use a multifaceted systems level approach to identify targets within innate immune signaling pathways that can safely provide disease modifying effects in AD. Comprehensive, transcriptomic, genetic and pathological data from both humans and mouse models will be generated, integrated and analyzed in novel ways. This integrated data will then be used to guide multiple preclinical target validation studies of key innate immune targets in both APP and tau mouse models as well as non-transgenic mice. These studies will dramatically accelerate the identification and validation of disease modifying innate immune modulatory strategies in AD and will provide important insights into how these various manipulations of innate immune activation states alter normal behaviors with an emphasis on cognition.

DESCRIPTION (provided by applicant): As part of an international collaboration, we provided strong evidence implicating TREM2 in LOAD by showing that TREM2 R47H is significantly associated with LOAD. In our combined case-control series of over 11,000 subjects, R47H has an odds ratio of 4.0 (P=6.6x10-9) similar to that of the well-known APOE ¿4 allele. In this series, a second TREM2 missense variant, D87N, also shows significant association. Remarkably, a recent study in which we participated shows that the risk associated with R47H extends beyond AD to FTD and PD, indicating that TREM2 may be a general neurodegenerative disease risk gene. TREM2 is a transmembrane signaling receptor in microglia known to function with its adaptor protein TYROBP (also known as DAP12) to effect non-inflammatory phagocytosis of apoptotic neurons. The disease-associated TREM2 variants may well act by altering the normal functioning of this transmembrane receptor. It is known, however, that soluble forms of TREM2 (sTREM2) are detectable in CSF, and are increased in patients with multiple sclerosis and CNS inflammation. The R47H and D87N variants are located in the extracellular domain of TREM2 and will, therefore, be present in sTREM2. Here, we explore the hypothesis that the disease-associated R47H and D87N variants may exert some or all of their disease-modifying effects on sTREM2 rather than on full-length TREM2 (FL-TREM2). sTREM2 could be generated by two mechanisms that are not mutually exclusive: proteolytic cleavage of FL-TREM2 and alternative splicing of exon 4, which contains the membrane-spanning domain. Our preliminary data and a recently published study show that, in cultured cells, sTREM2 is produced by proteolytic cleavage. R47H and D87N could act by altering this proteolytic cleavage thereby altering the relative amount of soluble and FL-TREM2. If sTREM2 functions as a decoy receptor for ligands that trigger signaling by FL-TREM2, then R47H and D87N may act additionally by altering the interaction of sTREM2 with these signaling ligands. To evaluate these mechanisms in vitro, we will quantitate sTREM2 and full-length TREM2 by sandwich ELISA in cell lines and in primary mouse microglia cultures expressing human WT, R47H, and D87N TREM2. To evaluate functional effects of sTREM2, we will generate soluble forms of WT, R47H, and D87N TREM2, and test their effects in functional assays analyzing direct binding to and phagocytosis of apoptotic neurons. In complementary in vivo studies, we will employ samples from the Mayo Clinic Study of Aging to evaluate associations (i) between sTREM2 and the R47H variant and (ii) between sTREM2 and conversion to MCI or AD. In brain samples, we will evaluate the association of R47H with sTREM2, FL- TREM2, the sTREM2/FL-TREM2 ratio, Braak stage, biochemical measures of A¿, and immunohistopathological measures of amyloid and tau pathology. We will also analyze sTREM2, FL-TREM2 and the sTREM2/FL-TREM2 ratio in AD as compared to non-AD brains.

? DESCRIPTION (provided by applicant): Genome-wide association studies (GWAS) in late-onset Alzheimer's disease (LOAD) identified candidate genetic loci, however cannot unequivocally uncover the disease gene or variants. Further, GWAS variants do not explain the full disease heritability. An important factor underlying this missing heritability may be rare functional disease variants of larger effect size that are missed by GWAS. Indeed, the recent identification of rare and strong AD risk variants in TREM2 via sequencing supports this hypothesis. Consequently, there are efforts for next-generation sequencing (NGS) in LOAD, however NGS comes with its own set of challenges. First, the large number of variants identified from NGS will require prioritization for downstream replication and functional studies. Second, appropriate assays are needed to test the functional consequences of these variants. Finally, NGS of large number of samples are still cost-prohibitive, precluding rapid functional assessment of variants. In this exploratory R21, we propose a cost-effective, novel alternative approach and plan to apply it to two intriguing nested candidate AD genes: CTNNA3 and LRRTM3. We will take advantage of existing and publicly available whole exome and genome sequence (WES, WGS) data to identify variants to test in our case-control cohort of ~11,000 subjects (Aim 1) and to assess the most promising, prioritized variants by in-vitro functional assays (Aim 2). Our preliminary data on the AD GWAS gene ABCA7, generated by this novel paradigm, as well as TREM2 findings, provide strong support for the feasibility of our approach. CTNNA3 and LRRTM3 reside in a linkage region of AD risk and amyloid ß levels identified independently by others and the co-PIs. They have opposite transcriptional orientation and strongly correlated gene expression levels in our published data, suggesting functional interactions. Both genes are implicated in synaptogenesis. We identified intronic variants in CTNNA3 that account for our Aß linkage signal. LRRTM3 influences APP processing. We showed protein interactions between LRRTM3, APP and BACE1 and herein demonstrate a role in long term potentiation. In summary, the nested CTNNA3/LRRTM3 are excellent candidate AD genes with potential roles in APP processing and/or synaptic physiology. We and others reported association of variants in both genes with AD risk, but replication has been inconsistent, which may in part be due to the lack of an assessment of rare variants. Our aims are: 1) To test rare variants in CTNNA3/LRRTM3 identified from 3 NIH funded NGS data, EVS (WES 4300 Caucasians, 2203 African-Americans), ADNI (WGS 246 Controls, 359 mild cognitive impairment, 184 AD), ADSP (584 subjects from 111 AD families), for AD risk association in our 11,000 subjects. 2) To test putative functional variants in CTNNA3/LRRTM3 in APP processing, cell toxicity and synaptic integrity. This exploratory R21 will enable a thorough and hypothesis-based study of two intriguing candidate AD genes and will also establish our cost-effective approach as a paradigm-shifting alternative to assess the deluge of genes being nominated in AD and other complex diseases.