Like the kid on a long car trip crowing, “Are we there yet?” from the back seat, many a researcher has surely wondered the same about the bumpy road to discovery of Alzheimer disease risk genes. Technological advances that have enabled genomewide association (GWA) analyses only intensify such ponderings. “These are early days,” said Rudy Tanzi of Massachusetts General Hospital, Boston, who led the latest GWA study in AD, published online yesterday in the American Journal of Human Genetics. “We have to wait until we have lots of samples tested on lots of genomewide screens.” In the meantime, using a first-ever family-based GWA approach to late-onset AD, his team presents several new risk marker candidates. “We found, beyond APOE, four novel genetic risk factors for Alzheimer’s that we believe have a strong potential for holding up as bona fide genetic hits because of the way this study was carried out,” Tanzi told ARF.

Previous genomewide screens in AD (Reiman et al., 2007; Grupe et al., 2007; Li et al., 2008) have been case-control studies that compare affected individuals with unrelated controls. In the new study, first authors Lars Bertram, also at Massachusetts General Hospital; Christoph Lange, Harvard School of Public Health, Boston; and colleagues performed a family-based GWA analysis. Using the Affymetrix GeneChip 500K array set—two arrays covering around 250,000 single-nucleotide polymorphisms (SNPs) each—they screened nearly 1,400 samples from 410 AD families collected as part of the National Institute of Mental Health (NIMH) Genetics Initiative Study. “In family-based studies, you compromise a little bit on power with same-size samples, but you generally get more rigorous and reliable results because you’re comparing one family at a time—comparing affected and unaffected siblings and looking for transmission,” Tanzi said. The other key aspect of this study that boosts its credibility was the fact that two of the four non-APOE genetic signals also showed significant association with AD in three independent samples totaling nearly 2,700 people from almost 900 additional families. “The big crux of genome association studies is that they often don’t replicate,” said Thomas Lehner, chief of the genomics research branch at NIMH. “Rudy made sure his study would not suffer that.”

The SNP showing the strongest association in this study appears to act, like APOE4, to modify onset age. Curiously, this SNP, rs11159647 resides on chromosome 14, about 10 million base pairs from the presenilin-1 (PS1) gene. The bad news: the marker falls within an intron of a predicted gene (NT_026437.1360) with unknown function. The predicted gene could be a transcription factor, Tanzi said, based on homology to a kruppel-like zinc-finger protein that appears to suppress expression of one of two presenilin homologs (HOP1) in nematodes. It remains unclear if this gene is related to the association with AD, however, since the authors found no evidence for linkage disequilibrium between rs11159647 and other SNPs in the kruppel-like zinc-finger homology region. They did find linkage disequilibrium between rs11159647 and SNPs in three flanking expressed sequence tags. These ESTs do not match the predicted exon sequences of the predicted zinc-finger protein, suggesting that there may be another gene, or genes, in the vicinity that are associated with AD. Interestingly, all three ESTs are expressed in the brain. The second strongest hit is found in the gene for CD33, a lectin involved in the innate immune system. The other two SNPs identified in this study occur within introns of a gene encoding a synaptic protein (BC040718) and the ATXN1 gene, which when mutated, can cause the neurodegenerative disease spinocerebellar ataxia.

None of these hits emerged as promising candidates in the previous AD genomewide screens, possibly due to differences in the statistical methods used. In case-control studies, p-values for each genetic association have to be corrected for the number of comparisons done, such that only the most robust signals (such as APOE) survive, Tanzi said. In the current study, his team used a two-stage approach: the first phase identified which SNPs had the most power, and subsequent analyses were restricted to that set of SNPs. With this method, Tanzi said, “the correction is much less penalizing.” Analyzing genotype data from two recently published AD GWA studies (Reiman et al., 2007; Li et al., 2008), his team did find nominally significant association for the strongest of the four SNPs in the Reiman et al. but not the Li et al. study. “That’s actually pretty darned good,” he said, “given the odds of seeing replication of a family-based study in a case-control study.”

Ultimately, the newly identified risk factors will need to pass muster by independent verification and analysis on Alzgene, a public database hosted by ARF and developed by Bertram, Tanzi, and others to provide a bioinformatics-based meta-analysis of AD genetic association studies. “In the end, we have to wait for labs A through Z to test them,” Tanzi said of the new markers. “They have to go through the same gauntlet as any candidate genes put out there. Replication and meta-analysis are the only things that will tell us which genes are real.”

A longstanding problem in the AD field has been the inability to replicate associations of small-effect candidate genes (see ARF related news story). A typical AD genetics study involving hundreds, or even several thousand, subjects lacks statistical power to reliably pull out such genes. The solution is larger datasets, said Alison Goate of Washington University in St. Louis, Missouri. “When you combine datasets so you have sample sizes in the range of 20,000 to 30,000, the genes of smaller effect fall out,” she told ARF, noting the success of similar efforts in cancer and diabetes. “You’re going to have a lot more power to see that these things of small effect are really significant.” Toward this end, the Alzheimer’s Disease Genetics Consortium (ADGC), which is headed by Gerald Schellenberg of the University of Pennsylvania in Philadelphia, and involves many AD genetics labs across the U.S., is working to combine existing data from multiple sites across multiple gene array platforms to enable analysis of a single large dataset.

Testing that many samples can be really expensive. At several hundred dollars per chip, the enormous cost of GWA studies may lead some to wonder whether they are truly worth the investment. To these skeptical folk, Tanzi simply points to the four genes currently known to either cause early onset AD (amyloid precursor protein [APP], PS1 and PS2) or increase risk for late-onset AD (APOE). “If we didn’t try to find the first genes, we would still be in the dark about what causes this disease at all,” he said. “The four genes we know about drive 99 percent of AD research.”

Buoyed by the discovery of these established factors, the field has focused heavily on the amyloid pathway as the key mechanism driving AD. But there is growing belief that other pathways are involved, perhaps involving these elusive smaller-effect genes. “You could imagine a scenario where in some individuals you could have a relatively small amyloid load and yet that leads to a degenerative process, whereas in other people you’d have a much larger amyloid load and still not have a degenerative process,” Goate said. “This would imply that there are other factors that influence whether or not you’re going to lose brain cells. Those would be important factors to know about.”

The search for new AD genes will undoubtedly suck up more research dollars, but the promise of what these efforts could deliver continues to drive Tanzi and others. “It’s expensive stuff, but unless we do it in many samples and do the meta-analysis, we’re never going to know what the rest of the disease genes are,” he said. “There’s no simple road.”—Esther Landhuis


No Available Comments

Make a Comment

To make a comment you must login or register.


News Citations

  1. Scouting AD’s “Little League” Genes: Can Systems Biology Help?

Paper Citations

  1. . GAB2 alleles modify Alzheimer's risk in APOE epsilon4 carriers. Neuron. 2007 Jun 7;54(5):713-20. PubMed.
  2. . Evidence for novel susceptibility genes for late-onset Alzheimer's disease from a genome-wide association study of putative functional variants. Hum Mol Genet. 2007 Apr 15;16(8):865-73. Epub 2007 Feb 22 PubMed.
  3. . Candidate single-nucleotide polymorphisms from a genomewide association study of Alzheimer disease. Arch Neurol. 2008 Jan;65(1):45-53. PubMed.

External Citations

  1. APOE
  2. BC040718
  3. Alzgene

Further Reading


  1. . Candidate single-nucleotide polymorphisms from a genomewide association study of Alzheimer disease. Arch Neurol. 2008 Jan;65(1):45-53. PubMed.
  2. . Evidence for novel susceptibility genes for late-onset Alzheimer's disease from a genome-wide association study of putative functional variants. Hum Mol Genet. 2007 Apr 15;16(8):865-73. Epub 2007 Feb 22 PubMed.

Primary Papers

  1. . Genome-wide association analysis reveals putative Alzheimer's disease susceptibility loci in addition to APOE. Am J Hum Genet. 2008 Nov;83(5):623-32. PubMed.