A plethora of pleiotropy across complex traits

A new analysis has identified hundreds of loci that are associated with multiple traits or diseases by comparing genome-wide association study (GWAS) data for 42 complex traits. The study uses the power of GWAS to provide evidence of pairs of traits with a likely causal relationship.

Pleiotropy is a phenomenon in genetics whereby a DNA mutation or variant has an effect on multiple traits. If traits are influenced by many genes of small effect (polygenicity) and pleiotropy is directional, then pleiotropy can be detected by the estimation of genetic correlation using pedigree data from traditional family studies1 or high-density molecular marker data from recent GWAS2,3. The genetic correlation parameter quantifies the overall proportion of genetic loci shared between traits, but it does not tell us where in the genome shared loci reside nor the underlying mechanism of the relationship between the traits. On page 709 of this issue, Joseph Pickrell and colleagues4, including the commercial company 23andMe, performed an analysis of 42 traits and tested for genetic loci that were associated with multiple traits, using summary-level GWAS data from the public domain and 23andMe. They identified 341 pleiotropic loci, which are likely to be only the tip of the iceberg. A number of variants were associated with phenotypes that are biologically not obviously associated: for example, the ABO gene locus was associated with coronary artery disease and tonsillectomy.

Plausible causal trait relationships

Pickrell et al. then asked whether the data could be used to infer causal relationships, in the sense that a genetic increase in one trait is always associated with a proportional increase in another trait, but not the other way around (Fig. 1). They report several such associations—for example, that body mass index (BMI) causally increases triglycerides: alleles that are associated with an increase in BMI always increase triglycerides by the same proportional amount, but alleles that increase triglycerides do not increase BMI. This is a very clever way to exploit GWAS data, in particular when the summary statistics are from traits that are not measured on the same people, so that environmental confounders are unlikely. It relies on GWAS for pairs of traits, each with multiple statistically highly significant loci, so that the effect sizes on each trait are estimated with little error. For such identified loci, the fact that effect sizes are very small does not matter given the extremely large sample sizes.

Figure 1: Inferring causal relationship between traits based on asymmetry in GWAS results.

(a,b) Shown is a hypothetical example in which trait X has a causal effect on trait Y. In this case, any genetic variant associated with X will appear to have an effect on Y, with the effect size proportional to that on X (a), whereas variants associated with Y do not show a corresponding effect on X (b). In this example, the SNPs identified from GWAS associated with trait X (a) are not the same as those identified from GWAS on trait Y (b).

The ubiquity of pleiotropy

In evolution, the strength of natural selection on a gene depends on the gene's pleiotropic effects on traits related to fitness5. In plant or animal breeding, directional pleiotropy is rampant because, when selection is applied to one trait, the mean of many other traits also changes over generations, sometimes in directions that not desirable for the breeder (for example, selecting for increased production may result in reduced reproduction). Selection experiments using laboratory species also observed this well over 50 years ago (and it was therefore somewhat puzzling that the mouse knockout community was surprised to observe that induced mutations often had effects on multiple phenotypes6). In human Mendelian (and clinical) genetics, pleiotropy is observed when a single mutation that causes disease can have multiple features. For example, gene mutations that cause intellectual disability can also result in facial dysmorphology and stunted growth. Indeed, the combination of multiple phenotypes from the same mutation is used to diagnose certain 'syndromes'. As an aside, such observations imply that we need to be careful in not being too 'trait-centric' when considering the role of genes or gene variants in biological processes. Another example is that the most common cause of cystic fibrosis (a disease mainly of the lung) also causes male infertility.

In human complex trait genetics, to date not much attention has been paid to the systematic study of pleiotropy7. This might be because the focus is often on the detection of genes or gene associations and a particular trait or disease of interest, or because multiple phenotypes are not usually available for the same samples. For example, for a case–control study of schizophrenia, it is unlikely that all cases and controls will be measured for multiple other phenotypes. Also relevant is that the research community first had to come to grips with the rampant polygenicity of complex traits (despite a few remaining sceptics, the empirical evidence that quantitative traits and common disease are highly polygenic is overwhelming), so that it was natural to start with individual traits and then move to the study of multiple traits.

One important contribution from Pickrell et al. is their asymmetry analysis (see Fig. 1) to distinguish between two different ways in which gene variants can be associated with multiple traits: that is, through independent biological pathways (for example, lung disease and infertility) or through one trait (or a genetic proxy) having a causative effect on a second trait (as with BMI and triglycerides). In human epidemiology, distinguishing between these two mechanisms is important, but in other applications, such as plant or animal breeding, it does not matter because the end result (a correlated response to selection) is the same.

Dissecting pleiotropy using GWAS

The availability of summary statistics from GWAS in the public domain has been tremendously helpful for researchers exploring pleiotropy, because the same traits do not have to be measured on the same individuals for new discoveries to be made. Inferring causality from genetic data on multiple traits, some of which are known to be risk factors for disease while others are known diseases or disorders, is exciting because it is much cheaper than performing many randomized controlled trials. Nevertheless, caution has to be applied when declaring traits to be causally related from statistical analyses of genetic data, because, as noted by Pickrell et al.4, there may be unobserved phenotypes that are genetically correlated with the observed 'causal' phenotypes that are the real cause.

There is also an increasing amount of data from GWAS on gene expression and DNA methylation (from expression quantitative trait locus (eQTL) and methylation quantitative trait locus (meQTL) studies), providing an opportunity to investigate pleiotropic effects (or genetic differences) in different tissues8. Testing pleiotropy by integrating the analysis of GWAS and eQTL data could help to pinpoint the most functionally relevant genes at GWAS loci for follow-up functional studies9. If multiple independent eQTLs or meQTLs are detected in relevant tissues, then the asymmetry analysis as performed by Pickrell et al. could be used to draw inferences about the causal effect of gene expression or DNA methylation to complex trait variation. This would be a remarkable advance in finding gene targets for complex diseases.

We can expect many more discoveries from multiple-trait GWAS analyses in the near future, including plausible claims of causality. The availability of GWAS data from large studies, in particular those with individual-level genetic data and multiple measured phenotypes, such as the UK Biobank, is essential to make new and likely unexpected discoveries. Phenome coverage will be more limiting than genome coverage. Our prediction is that it will become widely accepted that pleiotropy is the norm in genome–phenome associations.


  1. 1

    Lynch, M. & Walsh, B. Genetics and Analysis of Quantitative Traits (Sinauer Associates, Sunderland, Massachusetts, USA, 1998).

    Google Scholar 

  2. 2

    Lee, S.H. et al. Nat. Genet. 45, 984–994 (2013).

    CAS  Article  Google Scholar 

  3. 3

    Bulik-Sullivan, B. et al. Nat. Genet. 47, 1236–1241 (2015).

    CAS  Article  Google Scholar 

  4. 4

    Pickrell, J.K. et al. Nat. Genet. 48, 709–717 (2016).

    CAS  Article  Google Scholar 

  5. 5

    Williams, G.C. Evolution 11, 398–411 (1957).

    Article  Google Scholar 

  6. 6

    White, J.K. et al. Cell 154, 452–464 (2013).

    CAS  Article  Google Scholar 

  7. 7

    Solovieff, N., Cotsapas, C., Lee, P.H., Purcell, S.M. & Smoller, J.W. Nat. Rev. Genet. 14, 483–495 (2013).

    CAS  Article  Google Scholar 

  8. 8

    GTEx Consortium. Nat. Genet. 45, 580–585 (2013).

  9. 9

    Zhu, Z. et al. Nat. Genet. 48, 481–487 (2016).

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding authors

Correspondence to Peter M Visscher or Jian Yang.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Visscher, P., Yang, J. A plethora of pleiotropy across complex traits. Nat Genet 48, 707–708 (2016). https://doi.org/10.1038/ng.3604

Download citation

Further reading


Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing