Large genomewide association studies keep hatching new genetic variants associated with Alzheimer’s disease. To capitalize on this information, scientists need a way to quickly swat false positives and narrow in on the specific genes that truly affect disease risk. Functional screens that determine which gene candidates affect physical processes involved in Alzheimer’s could prove useful; however, many of these methods are expensive and time consuming. In a paper published online February 2 in the American Journal of Human Genetics, researchers led by Philip De Jager and Mel Feany at Brigham and Women’s Hospital in Boston, Massachusetts, describe a new approach to the problem. After performing a genomewide association study (GWAS) on a small cohort of 227 people, first author Joshua Shulman chose to screen promising gene candidates using a Drosophila tauopathy model. Out of 19 genes from 15 regions associated with risk for AD pathology, none of which had reached genomewide statistical significance, six interacted with tau pathology. The results of this pilot study suggest that the use of fly models has promise as a fast, low-cost, initial screen for GWAS hits.

“I think this is an excellent filter through which to push these long lists of results,” De Jager said, adding that the best candidates could then be studied further using models closer to the human disease, such as mouse lines, cell culture studies, or induced pluripotent stem cell lines.

Other scientists agree with the concept. “It’s a very clever approach,” said Minerva Carrasquillo at the Mayo Clinic in Jacksonville, Florida. “It’s going to be a good complementary approach to other methods, and it’s going to be informative.”

Large GWAS have recently turned up several AD risk genes, including CLU (clusterin/apolipoprotein J), CR1 (complement receptor 1), PICALM (phosphatidylinositol-binding clathrin assembly protein), and BIN1 (bridging integrator 1) (see ARF related news story on Harold et al., 2009 and Lambert et al., 2009; ARF related news story on Seshadri et al., 2010). A paper appearing this month in the Archives of General Psychiatry provides further replication for CLU, CR1, and PICALM in an independent dataset. In addition, the researchers, led by Lars Bertram at the Max Planck Institute for Molecular Genetics in Berlin, Germany, found that the AD risk allele of PICALM is associated with lower levels of Aβ42 in cerebrospinal fluid. Bertram oversees curation of the AlzGene database hosted on Alzforum.

GWAS are also finding genes that confer risk for other neurodegenerative diseases. In the February 1 Lancet, the International Parkinson Disease Genomics Consortium, led by Andrew Singleton at the National Institute on Aging, Bethesda, Maryland, reports finding 11 loci of genomewide significance for PD from a meta-analysis of five GWAS. Six had been previously identified; five were new.

GWAS also produce much longer lists of associations that fail to reach genomewide significance. More genes that interact with AD pathology are probably lurking among those loci. Scientists currently have limited options for fishing these genes out of the data. One option is to do even larger GWAS, or meta-analyses of several GWAS, and initiatives to do that in AD are underway. Another is to look for changes in the expression levels of candidate genes in people with AD. Neither method gives much insight into the functional role of gene candidates in the disease. Functional screens using mice or stem cells, however, are expensive and time consuming to develop.

“Larger GWAS are on the way,” Shulman points out. “We’re going to have increasing numbers of gene candidates to pursue.”

To see if Drosophila models could help process this flood of candidates, Shulman and colleagues carried out a small pilot study. They used a cohort of 227 participants in the Religious Orders Study and the Rush Memory and Aging Project, who had donated their brains after death. About 40 percent of the volunteers had AD, 40 percent were cognitively normal, and 20 percent had mild cognitive impairment. Rather than score each person by diagnosis, however, coauthor David Bennett and colleagues at Rush University in Chicago, Illinois, counted the number of amyloid plaques and neurofibrillary tangles in each brain. This provided a continuous measure of pathology, an endophenotype, which the authors then compared against the genetic data. The advantage of endophenotypes is that they supply more statistical power to detect associations because of the greater level of information available from each participant. The approach is creative, Carrasquillo said. She speculated that a GWAS of this small size would not have turned up any associations if the authors had simply scored each participant by diagnosis.

The study detected 24 independent single-nucleotide polymorphisms (SNPs) linked to AD pathology. Near 15 of those SNPs, the researchers found one or more candidate genes. The authors chose 19 genes that had orthologs in Drosophila for testing in a tauopathy model developed by Feany and colleagues (see ARF related news story on Wittmann et al., 2001). This model expresses mutant human tau and duplicates several features of AD, such as hyperphosphorylated tau and age-dependent neurodegeneration. Several pathways that mediate tau toxicity in this fly are active in mouse models and in humans, Feany said, which suggests the model is relevant to human disease. Tau expression in this fly produces small, roughened eyes, providing a visible phenotype for measuring the degree of tau toxicity.

Shulman and colleagues tested loss of function of each candidate gene using RNA interference. For some genes they were also able to examine gain of function, using Drosophila lines that increase gene expression. Six genes turned out to affect tau toxicity; in five, loss of function increased pathology. The top hit, GLUT14, encodes a glucose transporter. This fits with data showing disturbed glucose metabolism in AD. For example, less glucose transporter expression correlates with higher amounts of phosphorylated tau and neurofibrillary tangles (see Liu et al., 2008). The other genes include SLIT3, ELAVL2, β-spectrin, heparan sulfate 6-O-sulfotransferase, and discs large 1. Heparan sulfate proteoglycans have been shown to interact with SLITs, a family of axon guidance proteins (see e.g., Conway et al., 2011), suggesting that there could be functional interactions among some of these genes.

The Drosophila system has several advantages, Feany noted. It allows researchers to test genes in an in vivo setting, and capture functional interactions with other genes. Fly genetics are fast, inexpensive, and numerous well-characterized lines exist. Drosophila have been used as secondary screens for other mammalian systems, for example, in drug discovery (see ARF related news story on Karsten et al., 2006).

The system has limitations, too, the authors acknowledged. Genes of interest must have a Drosophila ortholog, for one thing. Carrasquillo said that tauopathy models would miss genes that directly influence APP processing or Aβ aggregation. Shulman and colleagues agree, and are in the process of testing their genes in Aβ transgenic flies. And while the fly screen can validate gene candidates, it cannot rule them out, De Jager said. “We have to be careful not to ignore interesting [GWAS] results simply because they aren’t validated in our method.”

In ongoing work, De Jager said, they are developing induced pluripotent stem cell lines of the gene variants for further functional investigation. Shulman said they also plan to take the pilot study into a much larger GWAS, using cognitive as well as pathological phenotypes. Other GWAS are interested in using the method, too. “We’ve begun collaborating with both the AD Genetic Consortium and the CHARGE Consortium for the purposes of applying the fly screen to their results,” Shulman said.

Shulman believes that applying a functional screen to GWAS data can help find genes associated with SNPs more effectively than using traditional statistical methods. Also, many AD-interacting genes may have too weak an effect to ever reach genomewide significance. “We’re going to need other methods to supplement the statistical approaches,” Shulman said. The authors see a Drosophila screen as but one tool in the geneticist’s toolkit, however. “You could envision a future where some list of GWAS [results] is rapidly screened through a number of different systems, maybe several fly transgenic models addressing tau or Aβ toxicity, cell culture models, and expression data sets,” said Shulman.—Madolyn Bowman Rogers

Comments

Make a Comment

To make a comment you must login or register.

Comments on News and Primary Papers

  1. In a report published in the American Journal of Human Genetics, Shulman and collaborators report a two-stage strategy to characterize new genetic determinants of Alzheimer’s disease. They first performed a genomewide association study (GWAS) in an autopsy cohort including 227 participants (91 AD cases, 50 MCI cases, and 86 controls). They selected their best hits for further analyses (p value less than 10-3).

    At this stage, they coupled this association study with a functional screening, with the postulate that convergence of both association and functional data might allow to restrict false-positive results, the main problem inherent to the genetic analyses, and to finally pick up relevant genes in the AD process.

    To perform this functional screening, they used a Drosophila model on the basis of in vivo interactions with the neurotoxicity of tau. They reported that SLC2A14 is a promising gene of interest, associated with AD risk. The Drosophila ortholog was associated with tau toxicity.

    The number of cases and controls analyzed in the GWAS step is limited, and replication/validation works are clearly needed. Unfortunately, this SNP was not associated with AD risk in our French GWAS (p = 0.76). However, the design of the two-stage strategy is clever and appears to be very powerful. The main advantages are as follows:

    • Most fundamental neuronal cell biological processes, such as synapse formation, neuronal communication, membrane trafficking, cell cycle regulation, and cell death, are very similar in Drosophila to those seen in humans. Hence, human neurodegenerative proteinopathy-mediated neurotoxicity can be successfully modeled in flies.
    • The high breeding rate and ease of handling and maintenance enables systematic functional genomic screening at a scale which is not conceivable in rodents. GWAS often define loci of interest encompassing several genes, and it can be difficult to determine which one is causal using only both genetic and literature bases. The systematic screening of all the genes within a locus can thus be a good option to finally pick up the relevant one.
    • Because GWAS are without hypothesis-driven strategies, an easy in vivo screening could provide a first attempt to understand how the genes are involved in the AD process.
    • To screen for genes exhibiting only suggestive associations can possibly highlight potential genetic determinants which would have been rejected only on statistical grounds (false negative).

    However, it is worth noting that this strategy also presents important limitations to keep in mind:

    • As mentioned by the authors, they only used a tau toxicity model and, of course, it is possible that the GWAS-defined genes are simply not involved in such a mechanism. Other complementary relevant Drosophila models might be of particular interest, e.g., Aβ toxicity models. As a consequence, it is not possible to determine whether a gene associated with AD risk is a false positive or not on the basis of this in vivo screening.
    • Even if 75 percent of the genes involved in human genetic disorders have at least one fly homologue (despite the considerable evolutionary distance between flies and humans), Drosophila could not allow for screening of all the GWAS-defined genes. For instance, no fly homologues have been reported for CLU and CR1, whereas clear homologues exist for BIN1 and PICALM.

    In conclusion, we think that this two-stage strategy will be particularly relevant when associated with a powerful GWAS. In this context, the recent announcement of the I-GAP consortium opens new perspectives with the possibility to screen for dozen of GWAS-defined genes/loci showing association reaching genomewide significant or suggestive p values.

    View all comments by Jean-Charles Lambert

References

News Citations

  1. Paper Alert: GWAS Hits Clusterin, CR1, PICALM Formally Published
  2. LOADing Up—Largest GWAS to Date Confirms Two, Adds Two Risk Genes
  3. In Fly Model of Tauopathy, Neurons Degenerate Without Tangles
  4. Paper Alert: Nipping Tau Tangles in the Bud

Paper Citations

  1. . Genome-wide association study identifies variants at CLU and PICALM associated with Alzheimer's disease. Nat Genet. 2009 Oct;41(10):1088-93. PubMed.
  2. . Genome-wide association study identifies variants at CLU and CR1 associated with Alzheimer's disease. Nat Genet. 2009 Oct;41(10):1094-9. PubMed.
  3. . Genome-wide analysis of genetic loci associated with Alzheimer disease. JAMA. 2010 May 12;303(18):1832-40. PubMed.
  4. . Tauopathy in Drosophila: neurodegeneration without neurofibrillary tangles. Science. 2001 Jul 27;293(5530):711-4. Epub 2001 Jun 14 PubMed.
  5. . Decreased glucose transporters correlate to abnormal hyperphosphorylation of tau in Alzheimer disease. FEBS Lett. 2008 Jan 23;582(2):359-64. PubMed.
  6. . Heparan sulfate sugar modifications mediate the functions of slits and other factors needed for mouse forebrain commissure development. J Neurosci. 2011 Feb 9;31(6):1955-70. PubMed.
  7. . A genomic screen for modifiers of tauopathy identifies puromycin-sensitive aminopeptidase as an inhibitor of tau-induced neurodegeneration. Neuron. 2006 Sep 7;51(5):549-60. PubMed.

External Citations

  1. CLU
  2. CR1
  3. PICALM
  4. BIN1
  5. AD Genetic Consortium
  6. CHARGE Consortium

Further Reading

Primary Papers

  1. . Functional screening of Alzheimer pathology genome-wide association signals in Drosophila. Am J Hum Genet. 2011 Feb 11;88(2):232-8. PubMed.
  2. . The role of clusterin, complement receptor 1, and phosphatidylinositol binding clathrin assembly protein in Alzheimer disease risk and cerebrospinal fluid biomarker levels. Arch Gen Psychiatry. 2011 Feb;68(2):207-13. PubMed.
  3. . Imputation of sequence variants for identification of genetic risks for Parkinson's disease: a meta-analysis of genome-wide association studies. Lancet. 2011 Feb 19;377(9766):641-9. PubMed.