DISCUSSION

The aim of this study is to test the influence of the position and orientation of Ac elements inserted at the maize p1 locus on the induction of homologous recombination in nearby sequences. The results indicate that Ac insertions in the 8.2-kbp interval between the 5.2-kbp direct repeats can significantly increase recombination frequencies when compared to insertion within or outside of the direct repeats. These results were initially observed in a somatic assay and were even more pronounced when germinal mutations were examined. Here, we discuss the implications of these results for proposed mechanisms of DSB repair and transposon-induced recombination in plants.

Homologous recombination initiated by a double-strand break: In yeast, it is well documented that DNA DSBs promote mitotic recombination (Orr-Weaveret al. 1981; Osmanet al. 1996). In plants, the frequency of intrachromosomal homologous recombination can be enhanced by X-ray, gamma-ray, and UV irradiation (Tovar and Lichtenstein 1992; Lebelet al. 1993; Puchtaet al. 1995) and by chemical agents that induce DSBs, such as methylmethanesulfonate and mitomycin C (Puchtaet al. 1995). Moreover, homologous recombination in plants is enhanced by in vivo induction of DNA double-strand breaks by a site-specific endonuclease (Puchtaet al. 1993; Chiurazziet al. 1996). In Drosophila, the P element can undergo precise excision in a homology-dependent process requiring P transposase (Engelset al. 1990). In Caenorhabditis elegans, the frequency of precise reversion following excision of the Tc1 element is much higher when heterozygous with a wild-type homologous chromosome (Plasterk 1991). Both of these examples indicate that DSBs generated by transposon excision can be repaired using the homologous locus as a template. The precise mechanism of Ac/Ds transposition is unknown, but genetic evidence indicates that it occurs via a “cut and paste” mechanism (Greenblatt and Brink 1962; Chenet al. 1992). Together, these studies are consistent with the hypothesis that transposon-induced homologous recombination is initiated by a DNA double-strand break generated by transposon excision.

Mechanism of transposon-induced recombination in plants: Ac insertion site and recombination frequency: A substrate with two direct repeats separated by an internal sequence was used to study DSB-induced intrachromosomal recombination in S. cerevisiae. The recombination frequency was >10 times higher when the DSB was induced in the internal sequence than at a site outside the repeats. In both cases, >90% of the recombinants were of the deletion type, in which the internal sequence was deleted but one direct repeat was retained (Prado and Aguilera 1995). This recombination product is similar to what we observed at the maize p1 locus. Similar results were obtained in the yeast Schizosaccharomyces pombe, where 99.8% recombinants were of the deletion type when the DSB was generated between the direct repeats, whereas if the DSBs were generated within one repeat, the deletions were reduced to 77% (Osmanet al. 1996). These deletions were proposed to occur via the single-strand annealing (SSA) pathway (Linet al. 1984). The SSA model proposes that the free 5′ ends produced by a DSB are degraded by a 5′ to 3′ exonuclease, exposing regions of homology on both sides of the break. The 3′ single strands anneal at complementary regions, and the protruding nonhomologous ends are removed by a flap endonuclease (Figure 4, steps c and d). According to this model, the reduction in deletion frequency observed when a DSB is induced within the repeat sequences reflects the reduced chance that a free 3′ single-stranded DNA end generated by 5′ to 3′ exonucleolytic digestion will find its complementary counterpart, due to the nonsymmetrical position of the DSB.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

—Models for transposon-induced homologous recombination. Red lines represent the inserted transposon; green lines, homologous direct repeats; thick black lines, sequences internal to the direct repeats; thin black lines, other genomic sequences. (a) Transposase binds to the terminal and subterminal regions of transposon causing the association of the transposon ends and generating a DNA double-strand break. (b) Exonuclease degrades from 5′ to 3′ at the DSB site and generates 3′ single-strand DNA. The transient DSB is repaired by one of three pathways: single-strand annealing (SSA; c and d); annealing-dependent synthesis (ADS; e-g); one-sided invasion (OSI; h-k). (c) The 5′ to 3′ degradation reaches the homologous region and two 3′ single-strand homologous sequences pair with each other. (d) Heterologous DNA is digested by nuclease and gaps are ligated. (e) Both strands of DSB are degraded by exonucleases that process the 5′ end faster than the 3′ end to form a 3′ overhang. (f and g) Once 3′ end degradation reaches homologous regions, complementary strands can anneal with each other to prime DNA synthesis and fill the remaining gaps. Green dotted lines indicate newly synthesized strands. (i-k) The 3′ end invades the homologous repeat copy. The heterologous DNA at the 3′ end is removed by nuclease. (j) D-loop is nicked at the red arrow, and DNA synthesis follows to generate a Holliday structure. (k) Resolution of the Holliday junction results in deletion of the p1 gene and retention of one copy of the direct repeat sequence.

In the case of the maize p1 locus, the interval between the two flanking repeats is ∼8.2 kbp. According to the SSA model, induction of a DSB at a site that is more centrally located within this interval would be expected to produce a higher frequency of recombination of the homologous flanking repeats compared to sites that are closer to one side of the interval, assuming that exonucleolytic degradation proceeds in both directions from the DSB at approximately equal rates. However, the recombination frequency of P1-ovov-Val is not significantly different from that of P1-ovov-12:1-1 and P1-ovov-1114 (Table 2), even though the locations of the Ac insertion alleles are very different: the Ac insertion site of allele P1-ovov-Val is 979 bp downstream of the end of the 5′ direct repeat of the p1 gene, while the Ac insertion sites of alleles P1-ovov-1114 and P1-ovov-12: 1-1 are 5.4 kb downstream of the end of the 5′ direct repeat of the p1 gene. Recombination in allele P1-ovov-Val via the SSA model would necessitate degradation of >12 kbp on the downstream side of the DSB to expose a complementary sequence for pairing with the upstream repeat.

A variation of the SSA model is termed synthesis-dependent strand annealing (SDSA) (Formosa and Alberts 1986). In this model, the free 3′ ends resulting from a DSB invade a homologous duplex and act as primers for DNA synthesis. The newly synthesized DNA strands anneal with each other, followed by further DNA synthesis to fill remaining gaps. On the basis of the SDSA model, we propose a model of annealing-dependent synthesis (ADS). Following formation of a DSB, both broken ends are degraded by exonucleases that process the 5′ end more rapidly than the 3′ end, resulting in 3′ overhanging strands on both sides of the break (Figure 4e). Once the 3′ end degradation reaches a region of homology (in the p1 locus within the flanking 5.2-kbp direct repeats), the complementary strands can anneal with each other, and the 3′ ends prime DNA synthesis to fill remaining gaps (Figure 4, f and g). This ADS model can account for the observed generation of the p1 mutants containing a single copy of the 5.2-kbp direct repeat sequence.

An alternative pathway for the repair of DSBs has been termed one-sided invasion (OSI; Belmaaza and Chartrand 1994); the OSI mechanism appears to be a major pathway in plant somatic recombination (Puchtaet al. 1996; Puchta 1998). To explain the formation of p1 gene deletions, the OSI model proposes that a free 3′ end generated by DSB would invade the opposite duplex. Complementary regions within the homologous repeat sequences would pair, and the nonhomologous 3′ end sequences would be removed by a flap endonuclease. Resolution of the resulting Holliday junction would result in deletion of the interval region and one repeat (Figure 4, h-k). A potential complication in the case of the p1 locus is that the invading 3′ ends may have several kilobase pairs of heterologous sequence arising from the interval between the direct repeats, and these heterologous sequences could affect the homology search and pairing with the downstream repeat sequences.

The behavior of an allele that contains Ac inserted within the direct repeat sequences provides a test of the predictions of the ADS and OSI models. Formation of a DSB by excision of Ac from the upstream direct repeat sequence would generate a free end homologous to the downstream direct repeat, and this should be the most efficient substrate for the OSI model, whereas the same free end would be a less efficient substrate for the ADS model: Exonucleolytic degradation in both directions from the DSB would likely result in deletion of the remaining ∼1 kbp portion of the upstream 5′ flanking repeat before the degradation of the downstream strand had proceeded the 12 kbp required to expose the complementary region in the downstream repeat sequence. The observed low frequency of deletions generated by the P1-9D47B allele is consistent with the expectations of the ADS model, but not the OSI model.

It should be noted that our ability to detect recombination events is based on a screen for loss of p1 expression; hence, we would not have detected gene conversion events that restore p1 function, if they did occur. Molecular analysis indicates that Ds elements may arise via double-strand gap repair following Ac transposition (Rubin and Levy 1997; Yanet al. 1999); however, such repair synthesis appears to be rare compared to the frequency of simple end-joining following Ac excision, regardless of whether homologous sequences are present in cis or in trans (Dooner and Martinez-Ferez 1997). Relevant to this study, Dooner and Martinez-Ferez (1997) observed no cases of conversion of the Ac insertion site by the adjacent duplication in the bz(Dp26)-m1 allele, which is structurally similar to P1-9D47B (i.e., a cis duplication with Ac inserted in one repeat copy). This suggests that gene conversion events that restore p1 function, if they occur, are probably too infrequent to account for the low frequency of recombination observed in the P1-9D47B allele.

How do transposons stimulate recombination? In addition to the importance of a DSB, transposon-induced recombination may involve the recruitment of host factors by transposase. Our data from transgenic Arabidopsis show that active transposons can greatly increase intrachromosomal homologous recombination (>1000-fold higher than control; Xiao and Peterson 2000). This frequency is 10-100 times higher than that observed for direct induction of a DSB (Chiurazziet al. 1996). One possible explanation for this high level of enhancement is that the efficiency of generating DSBs is different in these two experiments. Another explanation is that Ac transposase might recruit host factors that promote homologous recombination. There is some evidence that host plant factors are involved in transposition: First, host proteins that bind subterminal regions of Ac elements were found in both tobacco and maize (Becker and Kunze 1996; Levyet al. 1996). Second, a recessive mutation found in Arabidopsis can increase the frequency of Ac transposition (Jarviset al. 1997). Possibly, the Ac transposition process may recruit host factors required for a nonhomologous end-joining and/or homologous recombination. Following transposition, a DSB generated by transposon excision may be repaired by alternative pathways depending on the sequences or chromatin structure near the break (Scottet al. 1996). Further experiments will be required to elucidate the relative roles of DSBs, host factors, and chromatin structure in transposon-induced recombination.

The observation that Ac induces recombination between directly duplicated sequences contrasts with a recent report that Ac does not stimulate homologous meiotic recombination in the maize bz1 gene (Dooner and Martinez-Ferez 1997). To reconcile these apparently conflicting results, it is important to note that the transposon-induced recombination events we observed in maize (this report) and Arabidopsis (Xiao and Peterson 2000) are premeiotic. That is, these events occurred during sporophytic development to generate clonal sectors of genetically distinct cells. If these somatic sectors of mutant cells are included in the lineage that gives rise to the gametophytes, the premeiotic mutations can be transmitted to the next generation. In contrast, the events studied by Dooner and Martinez-Ferez (1997) appeared to occur at or near meiosis. The lack of effect of Ac excision on meiotic recombination may be due to temporal or spatial differences in the occurrence of Ac transposition relative to meiotic recombination. For example, Ac transposition is known to occur predominantly during or shortly after DNA replication (Greenblatt and Brink 1962; Chenet al. 1992), whereas meiotic recombination occurs after pairing of homologous chromosomes. Alternatively, Ac transposase may be spatially excluded from chromosomal regions that are undergoing meiotic recombination.

Anomalous mutation frequency of the P1-9D36A allele: Although the Ac insertion position in the P1-9D36A allele is between the two direct repeats, its colorless sector frequency is second lowest among all alleles. Nevertheless, the germinal recombination frequency of this allele is similar to that of other alleles with Ac inserted between the repeats. Thus, the P1-9D36A allele is suppressed in frequency of somatic mutations, yet it has a normal germinal mutation frequency. This mutation bias is opposite to that effected by a dominant mutation that suppresses germinal reversion of an allele of the maize waxy gene with a Ds insertion (Eisseset al. 1997). The observed low somatic mutation rate of the P1-9D36A allele could result from suppression of the activity of the Ac transposon. It has been demonstrated that the Ac element activity can vary depending on the Ac copy number and location (McClintock 1964; Heinlein 1995; for review see Fedoroff and Chandler 1994). Each of the alleles studied here was crossed to a r-sc:m-3 Ds tester stock; in this test, multiple copies of Ac elements delay the occurrence of Ds excision, resulting in small colored revertant sectors in the kernel aleurone (Kermicle 1980; Schwartz 1984). Ac activity of the P1-9D36A allele in the aleurone appeared normal; thus, suppression of Ac activity in pericarp cells, if it occurred, would have had to occur without affecting Ac activity in aleurone cells. Ac activity has also been shown to be inversely correlated with Ac methylation (Schwartz and Dennis 1986; Chometet al. 1987; Schwartz 1989; Brutnell and Dellaporta 1994). However, Southern hybridizations show that the BamHI site at the Ac 5′ end is unmethylated in the P1-9D36A allele (not shown); moreover, the Ac insertion in P1-9D36A is located in a region of the p1 gene that is relatively free of DNA methylation (Das and Messing 1994; Chopraet al. 1998). Therefore, the low frequency of colorless sectors in the P1-9D36A allele cannot be attributed to increased methylation of the Ac element, nor to the proximity of a hypermethylated region in the p1 locus. The basis for the anomalous somatic sector frequency of P1-9D36A remains to be determined.