Sensitive mapping of recombination hotspots using sequencing-based detection of ssDNA

Abstract

Meiotic DNA double-stranded breaks (DSBs) initiate genetic recombination in discrete areas of the genome called recombination hotspots. DSBs can be directly mapped using chromatin immunoprecipitation followed by sequencing (ChIP-seq). Nevertheless, the genome-wide mapping of recombination hotspots in mammals is still a challenge due to the low frequency of recombination, high heterogeneity of the germ cell population, and the relatively low efficiency of ChIP. To overcome these limitations we have developed a novel method--single-stranded DNA (ssDNA) sequencing (SSDS)--that specifically detects protein-bound single-stranded DNA at DSB ends. SSDS comprises a computational framework for the specific detection of ssDNA-derived reads in a sequencing library and a new library preparation procedure for the enrichment of fragments originating from ssDNA. The use of our technique reduces the nonspecific double-stranded DNA (dsDNA) background >10-fold. Our method can be extended to other systems where the identification of ssDNA or DSBs is desired.

Figure 1.

Detection of ssDNA-derived fragments in sequencing libraries. (A) Strategy for sequencing library preparation from single-stranded DNA fragments. (B) The proportion of ITR-containing reads (ITR length 6–20 nt) in α-DMC1, α-RAD51, and two control IgG samples. (C) The distribution of ITR lengths in α-DMC1 and α-RAD51 libraries. Frequency of sequencing read pairs having ITRs of a given length for all fragments (all reads) and type I ssDNA fragments (type I ssDNA) is plotted separately. (D) The length of micro-homology depends on the temperature of the reaction and on the end-repair enzyme.

Figure 2.

Enrichment of ssDNA-derived fragments in sequencing libraries. (A) Kinetic enrichment increases the proportion of ssDNA-derived reads in the sequencing libraries. The percentage of type I ssDNA- (ssDNA) and dsDNA-derived (dsDNA) reads in sequencing libraries prepared with (KE) or without (standard) kinetic enrichment. Antibody used in ChIP (α-DMC1 or α-RAD51) and genotypes (Psmc3ip^−/− or wt) of the mice used are indicated on the graph. (B) The distribution of ITR lengths in type I ssDNA fragments from two KE libraries. (C) Asymmetric sequencing largely corrects bias against longer ITRs. ITR length distribution for read pairs where the first read is uniquely mapped and the second read is not mapped using either 36 nt (36/36) or 40 nt (36/40) long second reads is plotted on the graph.

Figure 3.

Schematic representation of the kinetic enrichment approach. During the quick renaturation step the ssDNA-derived hairpins re-form almost immediately, generating ends suitable for adapter ligation (left). Complete renaturation of the dsDNA does not occur within this time frame, and the majority of the partially annealed products do not have ends suitable for ligation (right).

Figure 4.

Most of the DSB hotspot signal is ssDNA-specific. (A) ssDNA identification strongly reduces background in low-enrichment ChIP α-RAD51 Psmc3ip^−/−LE library. ssDNA + dsDNA or type I ssDNA only coverage profiles (reads per 1 kb, per million reads, RPKM) are plotted for a region on mouse chromosome 2. Coverage profiles were normalized to total number of ssDNA + dsDNA reads. α-RAD51 Psmc3ip^−/− LE library was prepared without using kinetic enrichment and ssDNA was computationally identified in the sequencing data. (B) The proportion of ssDNA-specific signal inside or outside all hotspots. While most of the signal inside hotspots is retained after ssDNA identification, outside of the hotspots, the background is reduced 10-fold or more. Sample specificity is plotted for α-DMC1 Psmc3ip^−/− (α-DMC1), α-RAD51 Psmc3ip^−/− (α-RAD51), and α-RAD51 Psmc3ip^−/− LE (α-RAD51 LE) samples. (C) Most of the signal inside hotspots originates from ssDNA. Mean depth of ssDNA and dsDNA read coverage (reads per 10 million reads) across all 9874 published mouse hotspots are plotted in 5-bp increments.

Figure 5.

Mapping meiotic DSB hotspots in wt mice using SSDS. (A) Meiotic DSB maps obtained using wt mice and regular ChIP-seq (α-DMC1 wt), Psmc3ip^−/− mice, and regular ChIP-seq (α-DMC1 Psmc3ip^−/−, reference data set; Smagulova et al. 2011) or using wt mice and SSDS (α-DMC1 wt KE) in a 2 Mb region of chromosome 1. We plot read coverage for all reads without ssDNA identification for regular ChIP-seq samples. Although peaks are located at the same places, hotspot detection is more sensitive in the KE library because of the lower background. Peaks are virtually undetectable in the wt sample prepared without kinetic enrichment. (B) Most hotspots are shared between the reference Psmc3ip^−/− data set and the wt KE samples. We detected hotspots in the α-DMC1 wt KE sample and compared their positions with the published hotspots (Smagulova et al. 2011). The number of peaks shared by and unique to these two data sets are plotted. (C) Hotspot detection in the wt KE sample is more specific. Mean depth of ssDNA and dsDNA read coverage (reads per 10 million reads) of Psmc3ip^−/− only (reference set) and KE only hotspots.