Fig 1.
Association of coronary artery disease (CAD) genetic risk and positive signatures of selection in 12 worldwide populations.
The 40 of 76 CAD genes investigated are shown that have at least four significant selection-risk associations in Panel B across all 12 populations. Panel A. Magnitude and significance of largest positive selection signal (integrated haplotype score, iHS) within each gene-population combination. P values (circles within squares) were obtained from 10000 permutations. Bonferroni corrected p value limit also shown (α = 0.05/76 = 0.000657) with closed circles. Panel B. Null hypothesis: no association between CAD genetic risk and positive selection, tested using mixed effects model with SNP estimates of CAD log odds genetic risk and iHS while accounting for gene LD structure as a random effect (first eigenvector from LD matrix per gene). Scaled regression coefficients were obtained directly from regressions, each p value from 10000 permutations. Panel C. Null hypothesis: association between genetic risk and positive selection for SNPs within CAD genes no different than non-CAD associated genes. Permuted p values were estimated by comparing each p value in Panel B against 100 nominal p values obtained by randomly choosing (without replacement) 100 non-CAD associated genes of similar size across the genome and using the same mixed effects model setup as described above. Populations. Grouped by ancestry, African (ASW, African ancestry in Southwest USA; MKK, Maasai in Kinyawa, Kenya; YRI, Yoruba from Ibadan, Nigeria; LWK, Luhya in Webuye, Kenya), East-Asian (CHB, Han Chinese subjects from Beijing; CHD, Chinese in Metropolitan Denver, Colorado; JPT, Japanese subjects from Tokyo), European (CEU, Utah residents with ancestry from northern and western Europe from the CEPH collection; TSI, Tuscans in Italy; FIN, Finnish in Finland), GIH (Gujarati Indians in Houston, TX, USA), MEX (Mexican ancestry in Los Angeles, CA, USA).
Table 1.
Leading multiple candidate selection signals in PHACTR1 SNPs.
Fig 2.
Quantitative links between coronary artery disease risk and selection signals in BCAS3.
A. Correlation between selection signals (iHS) and coronary artery disease (CAD) log odds genetic risk (log odds, ln(OR)), both represented as absolute values. Red line/upper right value, β from mixed effects regression. B. Base pair positional comparison of selection signals and CAD genetic risk across BCAS3. Blue points, CAD log odds values; grey-orange or non-significant-significant points, iHS scores. Horizontal bar shows BCAS3 gene (and intron) span and location of lead index SNP. Blue/orange lines are smoothed lines estimated with loess function in R. C. LD plots, r2. Populations: CEU, Utah residents with ancestry from northern and western Europe from the CEPH collection; YRI, Yoruba from Ibadan, Nigeria.
Fig 3.
Quantitative links between coronary artery disease risk and selection signals in PHACTR1.
A. Correlation between selection signals (iHS) and coronary artery disease (CAD) log odds genetic risk (ln(OR)), both represented as absolute values. Red line/upper right value, β from mixed effects regression. B. Base pair positional comparison of selection signals and CAD genetic risk across PHACTR1. Blue points, CAD log odds values; grey-orange or non-significant-significant points, iHS scores. Horizontal bar shows PHACTR1 gene (and intron) spans and location of index SNP if present. C. LD plots, r2. Populations: CEU, Utah residents with ancestry from northern and western Europe from the CEPH collection; GIH, Gujarati Indians in Houston, TX, USA.
Fig 4.
Comparing positive selection with gene regulation.
Summary distribution of permuted eQTL p values for SNPs with (left) or without (right) a significant selection signal. SNPs with a significant selection signal (iHS) were chosen by taking the largest significant positive selection signal (if one was present) within each gene-population combination. The same number of SNPs without a significant selection signal were also randomly drawn across all gene-population combinations for comparison. These SNPs were used in an eQTL analysis where they were regressed (including gender as a covariate) against their associated gene probe’s expression.
Fig 5.
Conceptual figure of potential evolutionary tradeoffs between coronary artery disease (CAD) burden and other phenotypes as a consequence of antagonistic pleiotropy (AP) [42].
As a simple example, AP describes gene effect on two traits (pleiotropy) that oppositely (antagonistic) affect individual fitness at different ages. Selection on that gene conferring a fitness advantage and disadvantage at different ages depends on the size and timing of the effects. An advantage during the ages with the highest probability of reproduction (between~20–45 years of age in humans) would increase fitness (lifetime reproductive success) more than a similarly sized disadvantage at later ages would decrease it. This concept is part of the well-known evolutionary theory of ageing, which describes tradeoffs in energy invested into growth, reproduction and survival [97]. In the figure above, intense natural selection occurring on CAD loci as a result of fitness advantages (+ signs, red text callout box 1.) conferred by genetically correlated risk factors (‘CAD risk factors’ box) or early-life traits (‘early-life traits’ box) trades off with the deleterious effects of these genes on fitness (i.e. CAD burden) later in life (- sign, red text callout box 2.) where the intensity of selection is weak. This occurs because of the negative relationship between genetic effects on early vs late-life traits (- sign, red text callout box 3.), which could help explain the high prevalence and maintenance of CAD in modern human populations. Over shorter timescales, lifetime probability of CAD is modified by a combination of genetic and environmental risk factors (e.g. [98]). There is evidence that such antagonistic effects have operated on CAD loci given: significant associations between CAD genetic risk and selection found (Figs 1 and 2); CAD genes are significantly enriched for lifetime reproductive success (S2 Table) and may also effect other early-life traits known to modify fitness (S4 Table); suggestive evidence was found for an antagonistic relationship between CAD and LRS (S3 Table); phenotypic selection has been found operating on CAD phenotypic risk factors [41].