Repair of DNA double-strand breaks by templated nucleotide sequence insertions derived from distant regions of the genome

To study the repair of DNA DSBs in vivo, we previously generated a vector (EF1aTK) containing the EF1a promoter driving expression of the herpes simplex thymidine kinase (HsTK); inserted between the EF1a promoter and HsTK cDNA was the recognition site for the rare-cutting meganuclease I-SceI (Fig. S1A). This vector was electroporated into the human monocytic leukemia cell line U937, and a clone (designated “F5”) that had integrated a single copy of the EF1aTK vector was isolated (Fig. 1A) (11). We attempted to induce reciprocal chromosomal translocations by introducing an I-SceI expression vector and using ganciclovir (GCV) to select clones that had lost expression of HsTK. The majority of the GCV-resistant clones identified had short deletions encompassing the HsTK start codon; however, we identified rare clones that had undergone an insertion of 47–756 bp that was not derived from nearby genomic sequences but instead was derived from a distant genomic region (11).

To gain a greater understanding of these insertions, we transfected the F5 clone with an I-SceI expression vector, harvested genomic DNA, and amplified the region flanking the I-SceI site. The PCR products then were subcloned into a plasmid vector, and individual colonies were isolated (Fig. 1A). We analyzed 120 samples; 10 (8.3%) were wild-type with an intact I-SceI site, 101 (84.2%) showed small (<50 bp) insertions or deletions (collectively designated “short indels”), 7 (5.8%) showed large (>50 bp) deletions, and 2 (1.7%) showed large (>50 bp) insertions.

Because insertion events were uncommon compared with short indels, we modified our experimental procedure by isolating a population of PCR products 50- to 1,500-bp larger than the wild-type EF1aTK allele to enrich for samples with insertions (SI Materials and Methods and Fig. 1A). With this modification, the frequency of PCR products containing insertions was enriched substantially (from 1.8 to 7–18%), although many small indels (78–86%) and large deletions (2–7%) still were recovered. We identified 32 insertions derived from distant genomic regions in the F5 cells. To determine whether this phenomenon was unique to the F5 clone, we repeated the experiment with a clone (designated “A15”) that had integrated a single copy of the EF1aTK vector into the ovarian cancer cell line OVCAR8 and again identified insertions derived from distant genomic regions. In all, 32 insertions from the F5 clone and 34 from the A15 clone were derived from distant regions of the genome (Dataset S1A). To distinguish these insertions from random, nontemplated nucleotide additions, we termed these insertions “templated-sequence insertions” (TSIs). One TSI from F5 cells and two TSIs from A15 cells were generated by the insertion of two unrelated fragments from distinct genomic origins. The lengths and distributions of the TSIs were similar in the two cell lines (P = 0.442, Student’s t test) (Fig. 1B, Fig. S1B, and Dataset S1A). The TSI origins were mapped to 18 of 24 human chromosomes without any obvious preference for particular chromosomes or chromosomal regions (Fig. S1B). More than half of the TSIs were derived from genic sequences (Fig. 1C), and approximately one fifth of the TSIs were templated from LINE-1 or short interspersed element (SINE) sequences (Fig. 1D). Junctions often showed features of aNHEJ, such as microhomology and nontemplated nucleotide addition (Fig. S1C). These findings demonstrate that the use of TSIs to patch I-SceI–induced DNA DSBs is a reproducible form of DNA-DSB repair.

Transcription activator-like effector nucleases (TALENs) are broadly applicable genome-editing tools (22) that produce site-specific DNA DSBs through the action of the FokI endonuclease (Fig. S1D). To determine whether repair of DNA DSBs by TSI is unique to I-SceI–induced DNA DSBs, we used TALENs to create site-specific DNA DSBs within the breast cancer 1 (BRCA1) and phosphatase and tensin homolog (PTEN) loci. Plasmids encoding TALENs for BRCA1 or PTEN were transfected into 293T or F5 cells, respectively. Genomic DNA was harvested and amplified with primers that flanked the TALEN cleavage site (Fig. S1D). Similar to the approach used in Fig. 1A, a region of the gel 50- to 1,500-bp larger than the wild-type, uncleaved fragment was excised and subcloned. Individual plasmids were sequenced, and, as shown in Fig. S1E, TALEN-induced DNA DSBs also were repaired by TSIs.

In addition to TSIs identified at experimentally induced DNA DSBs, there are rare, anecdotal reports consistent with insertions at the site of a physiologic DNA DSB, such as the IGH switch region in human B-lineage tumors (23–25); these findings are summarized in Dataset S2A. These observations support the hypothesis that the repair of DNA DSBs by TSIs is a generalizable phenomenon not restricted to repair of experimental DNA DSBs produced by endonucleases.

To determine whether the TSI fragment had been excised from the genome and used to patch a DNA DSB, we evaluated four pure daughter clones (derived from the F5 or A15 cell lines) which contained experimentally induced TSIs. We generated primers that flanked the TSI donor sequence and amplified genomic DNA from the specific clone to determine if the donor sequence had been deleted. All four daughter clones showed a single, specific PCR product identical in size to the parental cell line, suggesting that the TSI donor region had not been excised from the genome (Fig. S2A). It remained possible that a TSI donor allele had deleted more than 2 kb (a portion of which was inserted at the DNA-DSB site) and that the PCR product was amplified from the remaining intact allele. However, we identified an SNP within the 2-kb amplicon for clone 5T121; the nucleotide sequence of this SNP demonstrated that both alleles were retained (Fig. S2B). In addition, in all four examples, quantitative PCR (qPCR) showed a copy number gain for the templated sequence, supporting the hypothesis that TSI donor regions were not excised from the genome (Fig. S2C).

We hypothesized that the TSIs used to repair the DNA DSBs were patches produced by reverse transcription of a nuclear RNA template. To test the hypothesis that RNA could be used as a template for DNA-DSB repair, we cotransfected human F5 cells with an I-SceI expression vector and murine RNA, which had been treated with DNase I to eliminate any contaminating genomic DNA (Fig. S3 A and B), and isolated insertion events as described in Fig. 1A. We identified 102 clones with insertions; 92 clones had an insertion derived from a single donor site, nine clones had insertions derived from two donor sites, and one clone had an insertion derived from three distinct donor sites, leading to a total of 113 unique TSI events at the DNA-DSB site (Dataset S1B). Nine insertions matched murine sequences, suggesting that mRNA or pre-mRNA can be the underlying template for TSIs (Fig. 2). Seven of the nine mouse TSIs were derived from genic sequences, but two were derived from intergenic regions and would not have been expected to be present in total RNA. However, RT-PCR experiments demonstrated that these sequences represented nonannotated transcribed regions (Fig. S3 C and D). Unexpectedly, none of the insertions derived from murine RNA contained ribosomal RNA sequences, perhaps suggesting some specificity in the selection of the RNA template. Most (104/113) of the TSIs from the murine RNA cotransfection experiments were derived from endogenous human sequences; the size of the TSIs of human origin was 44–424 bp (Fig. S3E), similar in size and distribution to the TSIs identified in our initial experiments. The inserted sequences originated from 23 of the 24 human chromosomes with no clear preference for any chromosome (Fig. S3F). These results stand in contrast to insertions generated by replication fork stalling and strand switching, which preferentially insert nearby sequences derived from the same chromosome (26, 27). Of note, four insertions displayed simple repeat sequences, including three samples with mammalian telomere repeat (TTAGGG)n sequences, suggesting that telomerase RNA also can be used as a template to patch a DNA DSB (Fig. S3G).

LINEs are transposable elements that transpose not only LINE-1 RNA but also other elements such as SINEs (28). In addition, the reverse-transcriptase and endonuclease activity of the LINE-1 ORF2 has been proposed as a mechanism for the insertion of processed pseudogenes into the mammalian genome (29). To determine whether suppression of endogenous reverse-transcriptase enzymes would reduce the frequency of DNA-DSB repair by TSI, we transfected the F5 clone with an episomal I-SceI expression vector and selected cells that had been successfully transfected with hygromycin. The culture was divided into two aliquots. One was treated with AZT and 2′, 3′-dideoxyinosine (ddI), which have been shown to inhibit LINE-1 and other endogenous reverse-transcriptase enzymes (30, 31). The second aliquot was treated with vehicle alone. The concentrations of AZT and ddI used (3.7 and 4.2 μM, respectively) were below the AZT K_i for POLA, POLB, POLG, and POLE (140, 290, 8.7, and 400 μM, respectively) (32) and were below the IC₅₀ for muscle cell proliferation (100 μM for AZT and 500 μM for ddI) (33). Genomic DNA was harvested and amplified using primers that flank the I-SceI cleavage site. The region of the gel containing fragments 50- to 1,500-bp larger than the wild-type (uncleaved) fragment was excised, and the PCR products were subcloned. In each of four independent experiments, reverse-transcriptase inhibitor (RTI) treatment significantly reduced the frequency of TSIs after in vivo I-SceI cleavage (Fig. S4 A and B), supporting the hypothesis that the TSIs used to repair DNA DSBs are derived primarily by reverse transcription of RNA.

To determine whether TSIs derived from distant genomic regions were limited to repair of experimentally induced DNA DSBs or instead represented a recurrent mechanism for repair of a spontaneous DNA DSB, we studied whole-genome sequence data from two human myeloma cell lines, KP6 (34) and MC1286PE1. We reasoned that repair of a spontaneous DNA DSB by a TSI would produce pairs of structural variation (SV) in the whole-genome sequence, and therefore we examined SVs in these two cell lines. From a total of 3,393 interchromosomal SVs in the KP6 cell line and 3,145 SVs in the MC1286PE1 cell line, we identified pairs of reciprocal fusion sequences (compared with the reference genome, GRCh37/hg19) that could represent interchromosomal TSIs by applying the following criteria: The two fusion junctions had to be located within 50 kb of one another; the strand polarity had to align so that an insertion was feasible; and the sequences had to map to a single unique region of the genome. A detailed analysis of one of these SVs is shown in Fig. 3A. As outlined, although this pair of SVs could have been caused by a balanced translocation, the translocation would produce one dicentric chromosome and one acentric chromosome. We hypothesized that this pair of SVs instead was caused by an insertion of chromosome-4 sequences into chromosome 7, and we generated primers to test this hypothesis. As shown in Fig. 3 B and C, we were able to amplify this hypothetical TSI and verify that it indeed was produced by the insertion of chromosome-4 sequences into chromosome 7. Using this strategy, we verified 17 of 18 TSIs predicted from KP6 sequence data and 13 of 15 TSIs predicted from the MC1286PE1 sequence data (Fig. S5 and Dataset S2B).

Surprisingly, eight insertions were identical or nearly identical in the two cell lines (Dataset S2B), suggesting that these insertions represent polymorphisms in the human genome. To determine whether these TSIs were likely to be germ-line or somatic events, we compared the insertion junctions with a catalog of known SVs identified by the whole-genome sequence of 52 healthy individuals (SI Materials and Methods). SVs for 20 of the 23 unique TSIs were identified in this dataset, suggesting that most of the TSIs identified in the two myeloma cell lines represented germ-line polymorphisms rather than somatic mutations (Dataset S3A). To confirm the findings from the whole-genome SV dataset, we used PCR to amplify the TSIs from a set of widely studied human cell lines (293T, U937, HL60, K562, and OVCAR8). Nineteen of the 23 TSIs were PCR amplified from at least two independent cell lines, clearly demonstrating that these TSIs are polymorphic in human genomes (Dataset S3A); these polymorphic insertions are referred to as “templated sequence insertion polymorphisms” (TSIPs). Two TSIs were not found in the 52 individual databases or in cell lines but were found only in the KP6 cell line. These TSIs may represent somatic events, especially an 85-bp insertion of chromosome-1 sequence embedded at the junction of an acquired t(6;12) chromosomal translocation. Similar to our findings with the F5 and A15 daughter clones (Fig. S2), the TSI donor sequences from these two potential somatic TSIs were not deleted from the genome, and the copy number of the TSI donor sequence was increased (Fig. S6 A and B), as is consistent with the possibility of a somatic TSI. In addition, both TSI donor sequences were expressed in the KP6 cell line (Fig. S6C).

Features of the 23 unique TSIs identified in the two myeloma cell lines are shown in Dataset S3B. The TSIs can be placed into two classes (class 1 and class 2) based on nucleotide sequences at or near the insertion. Class 1 TSIs displayed a TSD of at least 5 bp and the addition of nontemplated adenines (a polyA “tail”) at one insertion junction. These class 1 TSIs typically inserted at a preferred LINE-1 retrotransposon integration site (consensus sequence 5′-TTTT/A-3′) and contained a polyadenylation signal (5′-AATAAA-3′) located 10–20 nucleotides upstream of the polyA track (Datasets S2B and S3B). Class 2 TSIs had none of these features but instead displayed aNHEJ features such as microdeletion, microhomology, and nontemplated nucleotide addition at the insertion sequence junction (Datasets S2B and S3B).

The TSI origin and acceptor loci are shown as a Circos plot in Fig. 4A. Twelve of the TSIs represent class 1 events, eight represent class 2 TSIs, and the remaining three are ambiguous, because they lack either a TSD or a polyA addition. Class 2 TSIs were shorter (85–305 bp; median, 164 bp) than class 1 TSIs (160–3771 bp; median, 464 bp) (P = 0.024, Student’s t test) (Fig. 4B) and were similar in size to our experimentally produced TSIs (Fig. 4B). Genic regions were overrepresented compared with genomic composition and were preferentially used as templates (Fig. 4C). We determined the frequency of the TSIPs identified in the two myeloma cell lines by examining SV data from the whole-genome sequences of 52 normal individuals (SI Materials and Methods). The prevalence of these TSIPs in the general population varied widely, from 0 to 98%, suggesting that both cell lines carried a unique combination of ancient, ancestral TSIPs and more recent, rare TSIPs (Fig. 4D).

The KP6 and MC1286PE1 cell lines have previously been published (34, 47). Data from the 52 normal volunteers was obtained by Complete Genomics and published on their publicly available web site (www.completegenomics.com/sequence-data/download-data/). Cell lines that contain a single copy of linearized pEF1aTK (11) (Fig. 1) integrated into the U937 human monocytic leukemia cell line (F5) (11) and the OVCAR-8 human ovary cancer cell line (A15) (12) have been described previously. An episomal I-SceI expression vector that contained a hygromycin-resistant cassette (pCEP4-I-SceI) (11) was transfected together with total murine RNA (20–120 μg) into F5 or A15 cells and was selected with hygromycin. Genomic DNA was isolated 14–21 d after transfection.

Genomic DNA was PCR amplified using primers that flanked the I-SceI site and was size-fractionated on 1% agarose gels. A portion of the gel 50- to 1,500-bp larger than the wild-type PCR product, predicted to contain unique polyclonal fragments with insertions, was excised, cloned into plasmid vectors, and transformed into bacteria. Colonies with insertions at the I-SceI cleavage site were sequenced.

Detailed experimental procedures are shown in SI Materials and Methods.