Repair-mediated duplication by capture of proximal chromosomal DNA has shaped vertebrate genome evolution

Abstract

DNA double-strand breaks (DSBs) are a common form of cellular damage that can lead to cell death if not repaired promptly. Experimental systems have shown that DSB repair in eukaryotic cells is often imperfect and may result in the insertion of extra chromosomal DNA or the duplication of existing DNA at the breakpoint. These events are thought to be a source of genomic instability and human diseases, but it is unclear whether they have contributed significantly to genome evolution. Here we developed an innovative computational pipeline that takes advantage of the repetitive structure of genomes to detect repair-mediated duplication events (RDs) that occurred in the germline and created insertions of at least 50 bp of genomic DNA. Using this pipeline we identified over 1,000 probable RDs in the human genome. Of these, 824 were intra-chromosomal, closely linked duplications of up to 619 bp bearing the hallmarks of the synthesis-dependent strand-annealing repair pathway. This mechanism has duplicated hundreds of sequences predicted to be functional in the human genome, including exons, UTRs, intron splice sites and transcription factor binding sites. Dating of the duplication events using comparative genomics and experimental validation revealed that the mechanism has operated continuously but with decreasing intensity throughout primate evolution. The mechanism has produced species-specific duplications in all primate species surveyed and is contributing to genomic variation among humans. Finally, we show that RDs have also occurred, albeit at a lower frequency, in non-primate mammals and other vertebrates, indicating that this mechanism has been an important force shaping vertebrate genome evolution.

Figure 1. SDSA DSB repair pathway.

(A) A DNA molecule suffers a double-strand break. The DSB can result in either blunt ends (left), 5′ overhanging ends (center), or 3′ overhanging ends (left). After the break occurs, exonuclease activity (dotted arrow) creates 3′ overhanging ends at the site of the lesion by removing nucleotides from blunt breaks (left) or breaks with 5′ overhanging ends (center). (B) After the exonuclease activity ceases, the resulting breakpoint has 3′ overhangs at both ends of the lesion. (C) In the SDSA pathway, one 3′ overhanging end can invade another DNA molecule, annealing to a sequence that is complementary, and repair synthesis begins. The other 3′ overhanging end may invade a different DNA molecule and begin repair synthesis as well. (D) After repair synthesis is completed and the new strand has dissociated from the template strand, it anneals with the other end of the initial lesion at a complementary region, creating the duplication of one template sequence (left). If both 3′ ends of the lesion used different template strands for repair synthesis, both strands will anneal at a complementary region, resulting in the duplication of two different template sequences (right).

Figure 2. Identification of interrupted TEs.

(A) Annotation of an uninterrupted TE (blue rectangle) along a chromosome. The chromosomal location is shown above the rectangle, while the positions of the TE consensus sequence are shown below. (B) Annotation of an interrupted TE. After the insertion of a new sequence (red rectangle), the two TE fragments are separated along the chromosome.

Figure 3. Chromosomal distance between donor and acceptor sequences located on the same chromosome in human.

Figure 4. RD where acceptor is a chimera of two different donor sequences.

Pre-insertion empty sites for the human-specific RD are shown for chimp and Rhesus macaque. The sequences for donor #1 and donor #2 are shown at the bottom, with a short stretch of base complementarity between the two donors in red, bold, and italics. The filler sequence is in blue, bold, and underlined.

Figure 5. Rate and timing of RD formation.

For each branch, the number of RDs created (n), the rate of repair-mediated duplication (per myr), and the average divergence of acceptor and donor pairs (%) during the time period are shown. The timing of RD formation for the human lineage (indicated by the red branches) is shown above the time scale by the red vertical bars. Each individual bar shows the relative proportion of RDs falling within the same, non-overlapping 5-myr bin.

Figure 6. PCR analysis of human-specific and chimpanzee-specific RD.

(A–D) PCR genotyping of the human-specific RD at chr15:31,987,740–31,988,005 in the 2006 hg18 assembly. For all 20 European (A) and 20 South American (B) individuals surveyed, the size of the PCR products is consistent with the homozygous presence of the acceptor sequence in all individuals tested. In contrast, among 20 African-American (C) and 20 Asian individuals (D) surveyed, several DNA samples yielded PCR products consistent with the absence of the acceptor sequence, either as homozygous [lanes 4–6 in (C) and 4, 9, and 10 in (D)] or heterozygous [lanes 3, 11, and 13 in (C) and 5, 7, and 8 in (D)]. See Table S2 for full loading orders. (E) Example of PCR validation of human-specific RD (chr10: 96,228,347–96,228,586): PCR products of size consistent with the presence of the acceptor sequence were obtained in human DNA (from HeLa cells or pooled individuals [Pop80]), while shorter PCR products, indicative of the absence of the acceptor sequence, were obtained with chimpanzee, gorilla, and orangutan DNA. (F) Example of PCR validation of chimpanzee-specific RD (chr7:136,854,307–136,854,636): PCR products of size consistent with the presence of the acceptor sequence were obtained with chimpanzee DNA (from single, “chimpanzee,” or pooled individuals [Pop12]), while shorter PCR products, indicative of the absence of the acceptor sequence, were obtained with human (HeLa cells), gorilla, and orangutan DNA.

Figure 7. Pre-integration empty sites in non-primate vertebrate species.

Microhomologies are in bold and italics, filler is in bold and underlined. Deletions are only underlined.