Distinct mutational behaviors differentiate short tandem repeats from microsatellites in the human genome

Abstract

A tandem repeat's (TR) propensity to mutate increases with repeat number, and can become very pronounced beyond a critical boundary, transforming it into a microsatellite (MS). However, a clear understanding of the mutational behavior of different TR classes and motifs and related mechanisms is lacking, as is a consensus on the existence of a boundary separating short TRs (STRs) from MSs. This hinders our understanding of MSs' mutational properties and their effective use as genetic markers. Using indel calls for 179 individuals from 1000 Genomes Pilot-1 Project, we determined polymorphism incidence for four major TR classes, and formalized its varying relationship with repeat number using segmented regression. We observed a biphasic regime with a transition from a faster to a slower exponential growth at 9, 5, 4, and 4 repeats for mono-, di-, tri-, and tetranucleotide TRs, respectively. We used an in vitro mutagenesis assay to evaluate the contribution of strand slippage errors to mutability. STRs and MSs differ in their absolute polymorphism levels, but more importantly in their rates of mutability growth. Although strand slippage is a major factor driving mononucleotide polymorphism incidence, dinucleotide polymorphism incidence is greater than that expected due to strand slippage alone, indicating that additional cellular factors might be driving dinucleotide mutability in the human genome. Leveraging on hundreds of human genomes, we present the first comprehensive, genome-wide analysis of TR mutational behavior, encompassing several motif sizes and compositions.

Fig. 1.—

East Asian (JPTCHB), European (CEU), and African (YRI) populations: polymorphism incidence (PI) curves (proportion of polymorphic TRs by increasing repeat number), separately for mono-, di-, tri-, and tetranucleotide TRs—on the log-scale.

Fig. 2.—

European (CEU) population: log of polymorphism incidence (see also fig. 1, black symbols) against repeat number, with fits from segmented regression (blue), for (A) mono-, (B) di-, (C) tri-, and (D) tetranucleotide TRs. Dotted vertical red lines show the location of the change points. Horizontal red lines represent 90% confidence intervals for change points. For mononucleotides, values at repeat number 2 and 3 were not included in the segmented regression fit due to convergence issue (see Materials and Methods for details).

Fig. 3.—

European (CEU) population: Polymorphism incidence curves (see also fig. 1, black symbols) and experimentally based Pol EF values (as fractions, red), both on log scale, for (A) mono- and (B) dinucleotide TRs.

Fig. 4.—

European (CEU) population: Polymorphism incidence curves (log scale) with bootstrap bands for (A) mononucleotide TRs and (B) dinucleotide TRs, separately for different motif composition. (C) Pol β unit-based indel error frequencies (log scale) with bootstrap bands, separately for different motif composition; data for GT and AC motifs are taken from (Kelkar et al. 2010). (D) Polymorphism incidence curves (log scale) with bootstrap bands for trinucleotide TRs, separately for different secondary structures.