17 Aug 2018

From the depths of more than 20,000 whole-exome sequences, researchers from the Alzheimer’s Disease Sequencing Project (ADSP) hauled out two novel genetic variants that significantly associate with risk for late-onset AD: the zinc-finger protein ZNF655, a transcription factor, and AC099552.4, a long noncoding RNA thought to regulate gene expression. A variant in a third gene—IGHG3, an immunoglobulin known to bind Aβ—missed statistical significance by a hair. The analysis, published August 14 in Molecular Psychiatry, also fingered several known AD risk genes, including TREM2, SORL1, and ABCA7. Led by Lindsay Farrer of Boston University, Adam Naj of the University of Pennsylvania in Philadelphia, and Myriam Fornage of the University of Texas Health Science Center in Houston, it represents the largest whole-exome sequencing (WES) study for AD to date.

Despite the growing breadth and depth of AD genetic studies, researchers estimate that a sizable chunk of the inherited risk remains unexplained. Genome-wide association studies (GWAS), which rely on extensive genotyping at single-nucleotide polymorphisms throughout the genome, are most useful for uncovering common genetic variants that each exert a teeny sway over disease, and these have identified more than 30 AD genes to date (Apr 2018 news). To search for rarer variants that might have larger effects, researchers have turned to whole-genome or whole-exome sequencing. These efforts have uncovered multiple rare variants, but are often constrained by small sample size (Aug 2017 news on Kunkle et al., 2017). 

The ADSP hopes to change that. The project, funded by the National Institutes of Health, has sequenced the exomes of some 11,000 people in a case-control study, and also sequenced the whole genomes of more than 600 people in family based cohorts (Beecham et al., 2017, for ADSP design details). A recent study led by Richard Mayeux of Columbia University in New York utilized the ADSP WES data set along with two other cohorts to hunt for rare loss-of-function variants linked to AD, ultimately finding only SORL1 (Jun 2018 news).

The current study is the first to use the ADSP WES data set as the sole source in its initial analysis, and it includes any type of variant associated with AD risk, as opposed to just the loss-of-function variety. To maximize the chances of finding novel genetic variants, ADSP case and control samples are selected based on risk scores. These are calculated using a combination of age, sex, ApoE genotype, age at AD onset (for cases), or age at last exam (for controls). Researchers then deem cases to be those with the highest risk scores that were unlikely to be explained by ApoE genotype alone, and controls as those with the lowest risk scores. Ultimately, this strategy whittled down the discovery cohort to 5,740 cases and 5,096 controls, about a third the size of the total ADSP cohort.

From these cases and controls, the authors identified more than 1.5 million single-nucleotide variants. Of these, 98 percent were exceedingly rare, cropping up in less than 5 percent of the cohort. To insure statistical power, co-first authors Joshua Bis, Xuiqiu Jian, Brian Kunkle, and Yuning Chen limited their search to the 160,898 variants that were carried by at least 10 participants in the cohort. From these, the researchers identified two well-known single variants that significantly associated with AD risk: the R47H variant in the microglial receptor TREM2, and a common variant in PILRA, an inhibitory immunoglobulin receptor expressed on myeloid cells. They also identified a novel rare variant in the long noncoding RNA AC099552.4. Fourteen other variants nominally associated with AD risk.

The researchers next moved up to gene level. They searched for genes containing more than one variant. Again, they limited the analysis to hits that occurred in at least 10 people and to variants expected to have a high impact on gene function. They excluded genes within 250kb of the ApoE locus to avoid any confound. With these constraints, they found two known AD genes (ABCA7 and TREM2) and two novel genes (OPRL1 and GAS2L2) that significantly associated with AD. Four other genes, including the zinc-finger protein ZNF655, trended toward an association.

Would OPRL1 and GAS2L2, or the long coding RNA from the single-variant analysis, replicate in other cohorts? To find out, the researchers ran meta-analyses on three additional AD case/control WES data sets: Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE), ADES-FR, and FinnAD. They also used the Alzheimer’s Disease Genetics Consortium GWAS data set. Although the GWAS data set did not include full exome sequences, the researchers used the Haplotype Reference Consortium to impute them. At the single-nucleotide level, they identified previously known AD variants in TREM2, MS4A6A, PILRA, and CR1. The researchers could not test the long noncoding RNA variant in this meta-analysis, because it was too rare in the replication data sets. However, in combining the discovery and replication meta-analyses, one other variant in the IGHG3 immunoglobulin gene came close to the threshold for an exome-wide significant effect on AD. At the gene level, ABCA7, as well as ZNF655, associated with AD.

Neither OPRL1 nor GAS2L2 appears to increase the odds of getting AD in the replication data sets. This comes as no surprise, Farrer pointed out, as those data sets were smaller, even when combined, than the discovery data set. Farrer said it was hard to do a good replication analysis because their discovery cohort is the biggest one available. Furthermore, the replication cohorts had not applied the same selection criteria—based on AD risk scores—to maximize the discovery of rare variants. In all, the researchers acknowledged that the replication analysis left something to be desired. Farrer told Alzforum that the ADSP will continue to recruit volunteers for DNA, and will eventually analyze double or triple the number of genomes reported in the current study.

In the heel of the hunt, AC099552.4 emerged from the discovery data set and ZNF655 from the replication data. Both implicate transcriptional regulation in AD. Zinc-finger proteins act as transcription factors, while long noncoding RNAs bind to target sequences in DNA to modulate transcription as well. What genes are controlled by these specific factors, and how they might instigate AD, remains to be explored.

Whether IGHG3 turns out to be an AD gene also remains to be seen. It encodes the constant (Fc) region of the IgGγ3 immunoglobulin heavy chain. Previous studies have reported that IgG antibodies can bind to Aβ aggregates, a finding that inspired the development of IVIG, a mixture of antibodies pooled from multiple donors that researchers hoped might clear Aβ from the brain, though the treatment failed to pass Phase 3 trials (see O'Nuallain  et al., 2008; May 2013 news). Three variants were identified in the gene, with varying levels of association with AD risk. How each affects gene function is also unclear. Two were missense mutations predicted to have a moderate effect on protein structure/function, while another was a synonymous variant. However, the researchers wrote that deletions in the gene have been found in more AD cases than controls in further analysis of the ADSP family cohort. Overall, the AD risk associated with this gene points to the involvement of the immune response in the etiology of the disease, Farrer said.—Jessica Shugart

Comments

No Available Comments

Make a Comment

To make a comment you must login or register.

References

News Citations

1. GWAS, GWAX: bioRχiv Hosts Bonanza of Alzheimer’s Genetics 26 Apr 2018
2. The Search for the Missing AD Heritability Turns Up New Rare Variants 15 Aug 2017
3. Gaining Notoriety, SORL1 Claims Spot Among Top Alzheimer’s Genes 7 Jun 2018
4. Gammagard™ Misses Endpoints in Phase 3 Trial 8 May 2013

Therapeutics Citations

1. Gammagard®

Paper Citations

1. Kunkle BW, Vardarajan BN, Naj AC, Whitehead PL, Rolati S, Slifer S, Carney RM, Cuccaro ML, Vance JM, Gilbert JR, Wang LS, Farrer LA, Reitz C, Haines JL, Beecham GW, Martin ER, Schellenberg GD, Mayeux RP, Pericak-Vance MA. Early-Onset Alzheimer Disease and Candidate Risk Genes Involved in Endolysosomal Transport. JAMA Neurol. 2017 Sep 1;74(9):1113-1122. PubMed.
2. Beecham GW, Bis JC, Martin ER, Choi SH, DeStefano AL, van Duijn CM, Fornage M, Gabriel SB, Koboldt DC, Larson DE, Naj AC, Psaty BM, Salerno W, Bush WS, Foroud TM, Wijsman E, Farrer LA, Goate A, Haines JL, Pericak-Vance MA, Boerwinkle E, Mayeux R, Seshadri S, Schellenberg G. The Alzheimer's Disease Sequencing Project: Study design and sample selection. Neurol Genet. 2017 Oct;3(5):e194. Epub 2017 Oct 13 PubMed.
3. O'Nuallain B, Acero L, Williams AD, Koeppen HP, Weber A, Schwarz HP, Wall JS, Weiss DT, Solomon A. Human plasma contains cross-reactive Abeta conformer-specific IgG antibodies. Biochemistry. 2008 Nov 25;47(47):12254-6. PubMed.

Further Reading

No Available Further Reading