Genome-wide analysis of microsatellite polymorphism in chicken circumventing the ascertainment bias

Abstract

Studies of microsatellites evolution based on marker data almost inherently suffer from an ascertainment bias because there is selection for the most mutable and polymorphic loci during marker development. To circumvent this bias we took advantage of whole-genome shotgun sequence data from three unrelated chicken individuals that, when aligned to the genome reference sequence, give sequence information on two chromosomes from about one-fourth (375,000) of all microsatellite loci containing di- through pentanucleotide repeat motifs in the chicken genome. Polymorphism is seen at loci with as few as five repeat units, and the proportion of dimorphic loci then increases to 50% for sequences with approximately 10 repeat units, to reach a maximum of 75%-80% for sequences with 15 or more repeat units. For any given repeat length, polymorphism increases with decreasing GC content of repeat motifs for dinucleotides, nonhairpin-forming trinucleotides, and tetranucleotides. For trinucleotide repeats which are likely to form hairpin structures, polymorphism increases with increasing GC content, indicating that the relative stability of hairpins affects the rate of replication slippage. For any given repeat length, polymorphism is significantly lower for imperfect compared to perfect repeats and repeat interruptions occur in >15% of loci. However, interruptions are not randomly distributed within repeat arrays but are preferentially located toward the ends. There is negative correlation between microsatellite abundance and single nucleotide polymorphism (SNP) density, providing large-scale genomic support for the hypothesis that equilibrium microsatellite distributions are governed by a balance between rate of replication slippage and rate of point mutation.

Figure 1.

Proportion of dimorphic loci in relation to repeat length for all microsatellites. Whiskers indicate the 95% confidence interval. Because of small sample size, results from the 15–20 and 20–35 repeat unit intervals have been pooled.

Figure 2.

Fitted logistic regression models of proportion polymorphic microsatellites as a function of microsatellite length for dinucleotide repeats (A), trinucleotide repeats (B), and tetranucleotide repeats (C). (A) (solid line) (AT)_n; (dashed line) (AC)_n and (AG)_n. (B) (Solid line) Motifs with 0% GC; (dashed line) 33% GC; (dotted line) 67% GC; (dotted-dashed line) 100% GC. (C) (Solid line) 0% GC motifs; (short dashed line) 25% GC; (dotted line) 50%; (dotted-dashed line) 75%; (long dashed line) 100% GC.

Figure 3.

Genomic occurrence of di- through pentanucleotide microsatellites in five vertebrates in relation to repeat length. (A) Dinucleotides; (B) trinucleotides; (C) tetranucleotides; (D) pentanucleotides.

Figure 4.

The relationship between SNP density and microsatetellite abundance in nonoverlapping 1 Mb windows (R² = −0.35, P = 10⁻¹⁵).