If one genome-wide association study yields a handful of new Alzheimer’s genes, then four GWAS combined should provide a double-handful or more. So reasoned the hundreds of researchers who joined forces in the International Genomics of Alzheimer’s Project (IGAP) meta-analysis to pick out genes whose effect size might have been too small to pop up in the individual GWAS. The researchers presented their findings at the Alzheimer’s Association International Conference, held in Boston July 13-18, and expect to publish them shortly. The work turned up 11 new loci that contribute to risk of late-onset AD, reported Jean-Charles Lambert of INSERM in Lille, France. Members of the collaboration are picking apart the GWAS data in various ways. For example, Peter Holmans of Cardiff University in the UK presented AD genomics through the lens of pathway analysis, hunting not for individual genes but for functionally related groups that contribute to risk. He picked out immunity, cholesterol metabolism, endocytosis, and ubiquitination as key modulators of AD.
“These studies are filling in the jigsaw of Alzheimer risk loci, and perhaps more importantly, they point to four pathways that are of interest, ”wrote John Hardy of University College London, UK, in an email to Alzforum (see full comment below). Rita Guerreiro, also at University College London, told Alzforum, “We have new genes that we want to follow, and that is always good, even if they have a very small effect.”
Scientists estimate that people inherit 60-80 percent of their risk for late-onset AD, with the remainder being environmental (see AlzGene). Previous GWAS identified nine loci; together with ApoE, these account for one-quarter of this heritability. That leaves plenty of genes still awaiting identification, and researchers formed IGAP in 2011 to narrow that breach. They pooled GWAS data—encompassing 17,000 AD cases and 38,000 controls—from four different studies. They are conducted by the European Alzheimer’s Disease Initiative, the UK-based Genetic and Environmental Risk in Alzheimer’s Disease collaboration, and two U.S. studies—the Cohorts for Heart and Aging Research in Genomic Epidemiology Consortium, and the Alzheimer’s Disease Genetics Consortium. They reasoned that the pooled data would give them greater ability to detect genetic variants with small effects on risk as well as rare variants.
The main IGAP analysis, presented by Lambert, included two stages. From the meta-analysis of the four previous GWAS studies, the researchers identified nearly 12,000 genetic loci potentially related to AD. In the second stage, they tested these with a new analysis of 8,500 cases and 11,000 controls. The result: 20 genes reached statistical significance. Of these, 11 were new hits pointing to loci listed below.
Lambert told Alzforum that the link between genes identified in an unbiased screen and the physiology of dementia is not always obvious at first glance, adding that plenty of research will be necessary to determine how each of these genes contributes to AD.
On the list of 20 genes, one was conspicuously absent—CD33. Though other GWAS have established its link to AD (see Hollingsworth et al., 2011 and Nai et al., 2011), CDC33 only reached statistical significance in the stage 1 IGAP analysis, but disappeared in stage 2. The authors carefully checked for technical issues, such as problems in genotyping or imputation, which corrects for missing data in a person's genotype, but could not explain why CD33 did not make the grade. Lambert said one possible explanation is that CD33 only plays a strong role in certain populations, which must not have been represented in the IGAP stage 2 sample. Both he and Guerreiro agreed that CD33’s absence here does not mean previous studies linking the gene to AD were incorrect. However, Guerreiro said, it does make her wonder what other AD genes have not yet surfaced in GWAS, emphasizing the importance of further studies.
Many genes may contribute to AD in ways too small to make a noticeable splash in the big pond of a GWAS. Holmans' pathway analysis approach asks whether gene variants that lack statistically significant linkage can still tell researchers something about AD risk. He sorted such genes from the IGAP analysis into different groups based on their involvement in similar biological functions. Then, he looked for gene-function categories that have more IGAP hits than would be expected by chance. Holmans has already applied this kind of analysis to bipolar disorder (Holmans et al., 2009) and Parkinson’s disease (Holmans et al., 2013).
Previously, the group reported a pathway study of AD using two published AD GWAS and at AAIC 2012, Holmans revealed that genes related to endocytosis and cholinergic receptors contribute to risk (see ARF related news story). In Boston, he presented results from the larger IGAP dataset. Four functional modules were enriched in this analysis: ubiquitination, endocytosis, immunity, and cholesterol metabolism. The study marks the first time that ubiquitination has popped up in an unbiased GWAS. None of the genes in that pathway came out in the individual GWAS; it took pathway analysis to identify their importance as a group.
“Peter’s talk was radically innovative,” commented David Brody of Washington University in St. Louis, Missouri. “We did not know that these four major pathways are heavily implicated in the heritability of Alzheimer’s disease.” Much of the value of Holmans’ analysis, he said, rests on GWAS being an entirely unbiased search and on the findings holding up in the new stage 2 sample.
Nonetheless, there are caveats to the pathway approach, Holmans himself said. For one, he was limited by the gene ontogeny categories available in published databases such as the Gene Ontogeny website and the Kyoto Encyclopedia of Genes and Genomes. If no one has created a particular category, then his analysis cannot identify that pathway. For example, APP processing did not come up, despite its link to AD. In addition, Holmans had to be careful that one powerful gene did not dominate a category. Because he was looking for groups of genes all with small influence, Holmans removed heavy hitters like ApoE from the analysis.
Holmans’ results support previous links between these pathways and AD (for example, see ARF related news story on Kurup et al., 2010 and Halawani et al., 2010; ARF related news story; ARF related news story; ARF related news story). They suggest that the pathways could be fruitful targets for therapeutics. Rather than targeting each individual gene in the group, he suggested scientists could look for a major player or biomarker to influence a whole pathway via new treatments.
IGAP scientists still have plenty to do, analyzing the dataset in different ways and determining how the identified genes relate to AD. At this point, Lambert said, it is difficult to calculate how much of AD heritability results from these 11 new genes. For one, the proper statistical tools to quantify their contribution to AD risk do not yet exist.
Lambert said that eventually GWAS will yield all the genes this approach can identify, which is mainly common variants with small effects. Already, scientists are turning to exome and whole-genome sequencing to identify rarer variants that may be strong risk factors, such as the recently identified Trem2 gene (see ARF related news story; ARF related news story; and ARF related news story).
The IGAP and other GWAS have shown scientists that polygenic diseases incorporate even more genetic risk factors than initially believed, commented Sudha Seshadri of Boston University, who co-chaired the IGAP session at the meeting. While this is perhaps bad news for the future prospect of personalized genomics, where doctors hope to identify disease risk based on genomes, it is good news for scientists aiming to develop treatments, she said. “We are expanding the number of biological targets to study,” Seshadri said.
New IGAP Genes
A region containing HLA-DRB5 and DRB1. These genes encode parts of the major histocompatibility complex II, which presents antigens on the surface of lymphocytes and macrophages. They influence susceptibility to multiple sclerosis (Sawcer et al., 2011) as well as Parkinson’s disease (Nalls et al., 2011).
SORL1 . This sortilin-related receptor is thought to function in endocytosis and cargo sorting. It has long been a suspect in AD (see ARF related news story; ARF related news story on Pottier et al., 2012; Miyashita et al., 2013), but had been controversial because not all studies were able to replicate the association. Now that it has showed up in IGAP, “SORL1 is clearly a good candidate,” Lambert concluded.
SLC24A4, solute carrier family 24, member 4. This gene encodes an ion exchanger involved in pigmentation of skin, hair and iris (Sulem et al., 2007, Duffy et al., 2010). It also appears to affect blood pressure and hypertension (Adeyemo et al., 2009). This role might be the link to AD, where cardiovascular function is a contributing factor, Lambert told Alzforum.
CASS4 codes for a scaffolding protein of unknown function.
INPP5D, inositol polyphosphate-5-phosphatase, dampens blood cell proliferation and survival.
MEF2C, myocyte enhancer factor 2C, encodes a transcription factor involved in muscle development. Mutations in the gene cause intellectual disability (Nowakowska et al., 2010, Le Meur et al., 2010). MEF2C has also been linked to synapstic plasticity (Akhtar et al., 2012).