30 March 2011. The common late-onset forms of Alzheimer’s and Parkinson’s disease are now believed to involve substantial inherited risk. With the use of increasingly large genomewide association studies (GWAS), scientists are making headway dissecting the genetic factors that contribute to these diseases, but GWAS results are merely lists of genetic markers that by themselves say little about the disease. GWAS are also limited to finding common variants, and can miss rare mutations that often carry more risk. At the 10th International Conference on Alzheimer’s and Parkinson’s Diseases, held 9-13 March 2011 in Barcelona, Spain, speakers outlined the field’s answers to these quandaries. They described how second-level analysis can identify disease-causing mutations from GWAS results, and suggested future strategies for finding variants that the current GWAS overlook. Their ultimate goal is to unearth the biology under each genetic association, the speakers emphasized, in order to provide new pathways for therapeutic intervention.
Finding the Disease Mutation
GWAS reveal associations between single nucleotide polymorphisms (SNPs) and disease risk, but they do not automatically point to a particular gene, much less identify the mutation that leads to the disease. “GWAS really identify genomic regions,” Andy Singleton at the National Institutes of Health, Bethesda, Maryland, wrote to ARF. “We associate a gene name with each locus for ease of tracking, but we are really not sure what the affected gene is.” The only way to be sure is to find a disease-causing mutation related to a particular gene. If the mutation is a protein coding (i.e., amino acid) change, then scientists can usually find it by sequencing only about 100 samples, said John Hardy at University College London, U.K. Most GWAS hits, however, involve non-coding changes, and these are much more difficult to spot, Hardy said. For example, neither clusterin (CLU) nor PICALM, two AD risk genes that are currently number 2 and 3 on AlzGene’s Top Results, includes simple coding changes. Non-coding variants typically affect mRNA splicing, expression levels, or stability, Hardy said. They may be caused by changes in gene promoter regions, the RNA tail, or in regulatory antisense or microRNAs. To help find these mutations, scientists at University College and the National Institutes of Health, Bethesda, Maryland, are developing a human brain expression database from more than 400 brains, which should be complete in about nine months, Hardy said. By providing information on normal gene expression levels in 10 brain regions, this database should help scientists spot changes in expression of GWAS hits in AD brains. Another promising technique, Hardy said, is to use transcriptome microarrays to find mRNA changes.
Christine Van Broeckhoven at the University of Antwerp, Belgium, described second-level analysis of several AD-associated GWAS hits. In particular, work published in the March 15 Molecular Psychiatry found that a splice variant in complement receptor 1 (CR1, place 6 on AlzGene) associates with AD risk. CR1 has two common isoforms, CR1-F and CR1-S. The less common CR1-S, with a population frequency of 15 percent, includes an extra copy of a repeat sequence that binds complement proteins C3b and C4b. In a Flanders-Belgian cohort of 1,039 patients and 844 controls, Van Broeckhoven and colleagues found that people with at least one copy of CR1-S had a 30 percent higher risk of AD than those with two copies of CR1-F. The scientists replicated the results in a French cohort, and confirmed by regression analysis that this copy number variation accounted for the original GWAS signal.
Van Broeckhoven pointed out that it is still not clear how CR1-S contributes to the disease. One possibility comes from the fact that when CR1 binds C3b, it inactivates C3b and inhibits the complement cascade. Therefore, the presence of an additional complement binding site on the receptor may dampen complement signaling and reduce Aβ clearance. However, this idea remains to be tested. Last week, a mouse genetics paper by Tony Wyss-Coray’s group at Stanford University implicated complement receptor in adult neurogenesis in the hippocampus (see ARF related news story).
In Barcelona, Van Broeckhoven also described preliminary findings on the GWAS hits CLU and bridging integrator 1 (BIN1). In a large Belgian population, her team found several rare variants that affect the β domain of CLU and were specific to AD patients, suggesting that β chain variations may be important in the disease. The researchers have replicated this result in a French cohort, and are in the process of looking at other samples. For BIN1, a protein involved in endocytosis and trafficking, preliminary results point to a genetic variant in the 5’ end of the protein, Van Broeckhoven said. This could indicate an effect on protein expression levels. The Belgian group also found an association between BIN1-associated SNPs and levels of tau in the cerebrospinal fluid.
In related news, a GWAS study led by Richard Mayeux at Columbia University, New York City, published in the March Archives of Neurology, replicated the original finding of CLU, PICALM, and BIN1 as AD risk factors in a Caribbean Hispanic population, as well as identifying some novel loci associated with late-onset AD.
Pinpointing the Biology
Even after finding a disease-causing mutation, the way in which that variant promotes AD poses the next puzzle. At that point, GWAS hits have to move into in vivo studies to find their biological effects, the speakers said. At AD/PD 2011, Peter St George-Hyslop at the University of Cambridge, U.K., described one such study. He performed genomewide microarrays on the brains of AD TgCRND8 mice at 70 days old, when Aβ levels first start to rise, as well as at 90 and 150 days, and compared gene expression levels with those of control mice. He replicated the result in J20 AD mice.
About 100 genes were dysregulated in AD brains, St George-Hyslop said. The genes clustered into just seven pathways, notably including inflammation, innate immunity, cholesterol, lipid metabolism, and protein trafficking. All of these pathways have been previously implicated in AD, and most GWAS hits also fall into these categories. The results suggest that AD genes act through Aβ-dependent mechanisms, St George-Hyslop said, perhaps by affecting how well the brain deals with Aβ. Aβ therapies may have to target whole pathways rather than particular genes to be successful, he suggested.
Hunting Down Missing Genetic Risk
Alzheimer’s has a strong inherited component, with estimates ranging from 60 to 80 percent, according to Van Broeckhoven. The established AD risk gene ApoE accounts for only a fraction of that variance, said Rudy Tanzi at Massachusetts General Hospital, Boston, and recent GWAS results have not made up the difference. The AlzGene database lists 42 gene candidates with significant meta-analysis results, Tanzi said, but each one has only a small effect on risk. The first-pass GWAS analysis may underestimate risk at each locus, however, said Tanzi, because rarer variants often exist that carry more risk (a point also emphasized by Hardy). As an example of this, Tanzi cited ADAM10, the “good” APP α-secretase. Currently on AlzGene spot 26, this gene was found to have two novel late-onset pro-domain mutations that segregated with the common SNP and increased AD risk (see ARF related news story on Kim et al., 2009).
Several speakers emphasized that GWAS alone cannot discover all disease variants. Large association studies are limited to finding common genetic variants that each contribute little risk. Traditional familial and linkage studies, on the other hand, uncover rare variants of high risk. In between these two extremes probably lie many low-frequency variants that carry moderate disease risk. The best way to find these genes, said many scientists, is to sequence exomes. That reveals rare coding variants. The exome consists of the parts of the genome that are translated into proteins. Since the exome makes up only about 1 percent of the total genome, it is more feasible to sequence exomes than whole genomes, but it is still expensive. Thomas Gasser at the Hertie-Institute of Clinical Brain Research, Tübingen, Germany, said that it currently costs about $5,000 to sequence one person’s exome, and this price may have to drop further for the technique to become widely used.
Exome Sequencing Ferrets Out a Rare PD Gene
Exome sequencing does work, however. In Barcelona, Carles Vilarino-Guell at the University of British Columbia, Vancouver, described the use of this technique to identify a rare PD mutation in a small Swiss family. The family has an autosomal-dominant form of PD, which on average began at age 51. By sequencing the exomes of two affected cousins, researchers identified 69 novel coding variants which the cousins had in common. Most of these variants were absent in other affected relatives or did not segregate with the disease, quickly narrowing the list of candidates to two. The scientists then pinned the disease-causing mutation to the Vps35 gene by comparing their results to a case-control series of more than 4,000 patients, in which the Vps35 mutation was seen only in people with PD, never in controls. In addition, the researchers have now found four different mutations in Vps35, all of which cause PD, in different families, Vilarino-Guell said.
Vps35 is part of the retromer complex, which sorts proteins from endosomes back to the trans-Golgi network for recycling. Vps35 has been implicated in Alzheimer’s disease (see ARF related news story and ARF news story). Sorting proteins have also turned up as culprits in frontotemporal dementia (see ARF related news story) and other neurodegenerative disorders (see ARF related news story). Vilarino-Guell suggested that the retromer complex may play an important role in several neurological diseases, but noted that the mechanism by which Vps35 mutations lead to PD still remains to be identified.
More New Risk Genes for Parkinson’s
Although Parkinson’s was once considered a sporadic disease, the last decade has turned up numerous genes involved in familial and early-onset PD, and scientists are now beginning to uncover candidate risk genes for the common late-onset cases as well. Singleton described a recently published meta-analysis of five GWAS conducted by the International Parkinson Disease Genomics Consortium over 18 months. The study included more than 5,000 people with PD and over 12,000 controls, and results were replicated in an independent set of more than 7,000 cases (see also ARF related news story). The data added to the evidence for six known risk gene candidates (SNCA; MAPT; LRRK2; HLA-DRA; GAK; BST1), as well as turning up five novel loci (associated with genes ACMSD; STK39; SYT11; MCCC1/LAMP3; and CCDC62/HIP1R). Interestingly, several of these genes, such as SNCA (α-synuclein) and LRRK2 (leucine-rich repeat kinase 2), also have variants that lead to early-onset PD. In AD, too, genetics started out with separate genes for early-onset (APP, presenilins) and late-onset cases (ApoE), but later presenilin mutations were found to be able to cause both forms of the disease. Gasser, who is also part of the PD consortium, said that the LRRK2 gene in particular has variants along the whole risk continuum from high to low, and that LRRK2 mutations are quite common, with a frequency up to 30 percent in some populations.
As is common in GWAS studies, the risk from each variant was small, but the cumulative effect was significant. The researchers divided the study population into five equal groups, from those who had the least number of risk variants to those who carried the most. Using risk profile analysis and comparing people with PD to controls, the scientists estimated that people in the highest fifth were 2.5 times more likely to have PD than those in the lowest fifth. Gasser estimates the 11 variants account for about 20 percent of the total genetic variance of PD, with familial PD genes contributing another 5 percent of genetic risk. That leaves a big genetic gap still to be filled by exome sequencing, Gasser noted.
A prominent example of a low-frequency, moderate-risk PD gene is the GBA gene encoding the metabolic enzyme glucocerebrosidase. It was recently discovered to have PD-associated mutations, and has already jumped to place 4 on PDGene Top Results. GBA did not show up in the GWAS results. Gasser said it is the type of gene that slips by GWAS because it has numerous individually rare variants. At AD/PD 2011, Núria Setó-Salvia, Hospital Sant Pau, Barcelona, and Jose Luis Capablo Liesa, University of Zaragoza, Spain, presented independent posters on GBA. They confirmed that sequencing of the whole gene, or at least exome sequencing that included exon-intron boundaries, was necessary to reveal known and new mutations in their respective samples of patients with PD and with dementia with Lewy bodies.—Madolyn Bowman Rogers.
Brouwers N, Van Cauwenberghe C, Engelborghs S, Lambert JC, Bettens K, Le Bastard N, Pasquier F, Montoya AG, Peeters K, Mattheijssens M, Vandenberghe R, De Deyn PP, Cruts M, Amouyel P, Sleegers K, Van Broeckhoven C. Alzheimer risk associated with a copy number variation in the complement receptor 1 increasing C3b/C4b binding sites. Mol Psychiatry. 2011 Mar 15. Abstract
Lee JH, Cheng R, Barral S, Reitz C, Medrano M, Lantigua R, Jiménez-Velazquez IZ, Rogaeva E, St George-Hyslop PH, Mayeux R. Identification of novel loci for Alzheimer disease and replication of CLU, PICALM, and BIN1 in Caribbean Hispanic individuals. Arch Neurol. 2011 Mar;68(3):320-8. Abstract