Negative selection on BRCA1 susceptibility alleles sheds light on the population genetics of late-onset diseases and aging theory

Abstract

The magnitude of negative selection on alleles involved in age-specific mortality decreases with age. This is the foundation of the evolutionary theory of senescence. Because of this decrease in negative selection with age, and because of the absence of reproduction after menopause, alleles involved in women's late-onset diseases are generally considered evolutionarily neutral. Recently, genetic and epidemiological data on alleles involved in late onset-diseases have become available. It is therefore timely to estimate selection on these alleles. Here, we estimate selection on BRCA1 alleles leading to susceptibility to late-onset breast and ovarian cancer. For this, we integrate estimates of the risk of developing a cancer for BRCA1-carriers into population genetics frameworks, and calculate selection coefficients on BRCA1 alleles for different demographic scenarios varying across the extent of human demography. We then explore the magnitude of negative selection on alleles leading to a diverse range of risk patterns, to capture a variety of late-onset diseases. We show that BRCA1 alleles may have been under significant negative selection during human history. Although the mean age of onset of the disease is long after menopause, variance in age of onset means that there are always enough cases occurring before the end of reproductive life to compromise the selective value of women carrying a susceptibility allele in BRCA1. This seems to be the case for an extended range of risk of onset functions varying both in mean and variance. This finding may explain the distribution of BRCA1 alleles' frequency, and also why alleles for many late-onset diseases, like certain familial forms of cancer, coronary artery diseases and Alzheimer dementia, are typically recent and rare. Finally, we discuss why the two most popular evolutionary theories of aging, mutation accumulation and antagonistic pleiotropy, may underestimate the effect of selection on survival at old ages.

Figure 1. Age-specific risk of developing the disease.

Women's cumulative risk of developing breast cancer (dashed line, F_b(x)), ovarian cancer (dotted line, F_o(x)) or either (solid line, F_bo(x)) in a population free of any other causes of death. The two functions F_b(x) and F_o(x) are fitted 2-parameter cumulative gamma distributions from data from . For breast cancer, F_b(x), α = 7.93 and β = 7.54; and for ovarian cancer, F_o(x), α = 8.20 and β = 9.86. The cumulative gamma function associated with this risk F_bo(x) has parameters α = 9.97 and β = 5.54.

Figure 2. Demographic scenarios.

Adult women's survival (A) and fertility (B) from ages 10 to 50 for the population of early Quebec (solid line), the Ache (dashed line) and the modeled population (dotted line). These populations differ by their growth (respectively high, intermediate and stationary, see [12], [13], [14]). Female adults' age-specific survivals were modeled by a Gompertz-Makeham. For Quebec, a = 4.69×10⁻⁴, b = 7.28×10⁻², Δ = 0 and c = 1.49×10⁻³. For the Ache, a = 3.00×10⁻⁵, b = 1.10×10⁻¹, Δ = 0 and c = 6.33×10⁻³. For the modeled population, a = 1.99×10⁻³, b = 5.82×10⁻², Δ = 10 and c = 1.2×10⁻². Female adults' fertilities were modeled by a Hadwiger function. For Quebec, TFR = 12.25, θ₁ = −246.92, θ₂ = 18.72, θ₃ = 272.77. For the Ache, TFR = 8.032, θ₁ = −36.82, θ₂ = 4.79, θ₃ = 68.29. For the modeled population TFR = 2.578, θ₁ = −7.61, θ₂ = 3.31, θ₃ = 34.31.

Figure 3. Selection on late-onset diseases.

To explore a range of patterns of age of onset, we altered the observed cumulative risk with age F_bo(x) of developing either breast or ovarian cancer for females carrying a BRCA1 susceptibility allele estimated from (red line) by changing the mean from age 40 to 70 (A) and the variance from 100 to 800 (B). This was achieved by altering parameters α and β of the cumulative gamma. We calculated the minimum effective population size Ne_min = 10/2s necessary for negative selection to dominate over drift corresponding to BRCA1-alleles associated with each of these curves. These Ne_min are shown in (C) for the Quebec population. The point represents values estimated from for the cumulative risk of developing breast and ovarian cancer given a mutation in BRCA1 (mean = 55; variance = 306; Ne_min = 92).

Figure 4. Selection on late-onset diseases for the Ache and the modeled population.

As in Figure 3C but using survival and fertility for Ache (left panel) and the modeled population (right panel).