5 Jun 2017
by Marina Chicurel
This article appears by special arrangement with Alzforum. See original article with additional links/commentary.
Genomewide association studies (GWAS) have pinpointed genetic variations associated with various human traits, even complex ones influenced by multiple genes, such as height. But finding robust ties to intelligence has proven elusive. By harnessing the statistical power of data from nearly 80,000 people, scientists led by Danielle Posthuma at VU University Amsterdam have now added 40 new intelligence-linked genes to the current short list. The results appear in the May 22 issue of Nature Genetics. Among the traits that also associate with these new variants, Alzheimer’s emerged as the most negatively correlated.
Based on studies of families, particularly twins, researchers have known for decades that intelligence is highly heritable (Jansen et al., 2015). But nailing down which genes are responsible has been challenging. Measuring intellectual ability precisely and reliably is a problem, but more importantly, most genetic studies have been too small to accurately uncover the subtle effects of individual gene variants and probably identified false positives as well, researchers say (Chabris et al., 2012).
Working in Posthuma’s lab, first author Suzanne Sniekers analyzed data from 78,308 individuals from 13 cohorts composed of children and adults of European descent. The sample included published GWAS results for intelligence on about 50,000 individuals (Benyamin et al., 2014). Sniekers correlated genetic variance with pooled data from a variety of IQ tests, most involving short, web-based quizzes taken remotely. “Luckily, there’s a high correlation between the results of the different tests, which we confirmed when we looked at the association data from each cohort independently,” said Posthuma.
Todd Lencz, from the Hofstra Northwell School of Medicine in Glen Oaks, New York, said he expected the short tests would be too crude. This is why his team stuck to the more rigorous tests in a recent study in which he identified two loci and seven genes associated with general cognitive function among 35,298 individuals (Trampush et al., 2017). “The proof is in the pudding, though,” said Lencz. “The short tests turned out to be robust enough.”
To search for associations, Sniekers used standard GWAS, as well as a variant known as genomewide gene association study (GWGAS), in which the effects of single-nucleotide polymorphisms (SNPs) within the same gene are combined to uncover hits that can be missed when the effects of individual SNPs fail to reach statistical significance.
The researchers found three loci and 12 genes previously associated with cognition, plus they identified 15 new loci and 40 new genes. Many of the 52 genes are involved in neuronal function (including DCC, APBA1, PRR7, ZFHX3, HCRTR1, NEGR1, MEF2C, SHANK3, and ATXN2L), with several implicated in neuronal cell development. The strongest association was with FOXO3, a gene involved in insulin signaling. It has been linked to longevity (Willcox et al., 2008). In humans, FOXO3 and APOE are the only two genes whose association with lifespan has been widely replicated (Broer et al., 2015).
Despite the 40 new intelligence genes, the list is far from complete, researchers say. Lencz anticipates that hundreds, if not thousands, of additional variants exist. As GWAS studies get larger, more of these will be identified. “It’s like prospecting for gold, you keep searching and accumulating more in your bin,” he said.
To extend the reach of their study, Posthuma’s team compared their data to GWAS data for 32 other traits. Consistent with previous studies, they found education level was highly correlated with intelligence. They also found moderate, positive correlations with 13 other traits, including intracranial volume, head circumference in infancy, and success quitting smoking.
Peter Visscher, an expert in the genetics of complex traits at the University of Queensland in Brisbane, was unsurprised by the results. “The high genetic correlation with education attainment has been reported before (Okbay et al., 2016),” he wrote. Looking ahead, he considered that studies of education level will be critical for expanding our understanding of intelligence. “Sample size is king and it is much easier to get huge sample sizes for education,” he said. “It seems likely that in the near future there will be GWAS studies on education on millions of individuals, and such studies are likely to detect many hundreds of loci in the genome.” Posthuma is looking forward to the UK Biobank releasing data sets for 500,000 individuals, up from their current public collection of 100,000.
Although the study didn’t focus on Alzheimer’s, some curious connections emerged. FOXO3’s role in insulin signaling might be relevant given that several studies have linked insulin resistance to AD pathogenesis (for review see de la Monte, 2012; AlzRisk). Also, two of the 52 genes associated with intelligence, MEF2C and EXOC4, have been linked to synaptic function and AD.
Ekaterina Rogaeva of the University of Toronto previously pinpointed a region in the EXOC4 locus that associates with AD in Caribbean Hispanics, who have a higher incidence of AD than many other populations (Ghani et al., 2013). Rogaeva wondered if Posthuma’s results would replicate in non-Europeans. AD also surfaced as the trait most negatively correlated with intelligence. This ties in with the idea that people with a lot of cognitive reserve may escape frank dementia, even when they have underlying AD pathology. (See comment below.)
Can one conclude, then, that smart people are less likely to get AD? No, says Carlos Cruchaga at Washington University, Saint Louis. The overlap may shed light on common pathways, but what the correlations mean clinically is unclear. “There is more overlap in complex traits, including many neurological disorders, than we expected, but it’s too early to say how it all works,” said Cruchaga.